Index: Makefile ================================================================== --- Makefile +++ Makefile @@ -183,14 +183,14 @@ mofiles/dbmod.o : mofiles/commonmod.o mofiles/keysmod.o \ mofiles/tasksmod.o mofiles/odsmod.o mofiles/testsmod.o mofiles/tasksmod.o : mofiles/commonmod.o mofiles/pgdbmod.o mofiles/rmtmod.o : mofiles/commonmod.o \ mofiles/apimod.o mofiles/ulex.o mofiles/itemsmod.o -mofiles/apimod.o : mofiles/dbmod.o mofiles/commonmod.o +mofiles/apimod.o : mofiles/dbmod.o mofiles/commonmod.o mofiles/servermod.o mofiles/runsmod.o : mofiles/testsmod.o mofiles/mtmod.o : mofiles/mtconfigf.o mofiles/rmtmod.o -mofiles/servermod.o : mofiles/commonmod.o +mofiles/servermod.o : mofiles/commonmod.o mofiles/dbmod.o # Removed from megamod.o dep: mofiles/ftail.o mofiles/megamod.o : \ mofiles/rmtmod.o \ mofiles/commonmod.o \ Index: apimod.scm ================================================================== --- apimod.scm +++ apimod.scm @@ -19,10 +19,11 @@ ;;====================================================================== (declare (unit apimod)) (declare (uses commonmod)) (declare (uses dbmod)) +(declare (uses servermod)) (module apimod * (import scheme chicken data-structures extras) @@ -29,10 +30,11 @@ (import (prefix sqlite3 sqlite3:) posix typed-records srfi-18 srfi-69 format ports srfi-1 matchable s11n z3 (prefix base64 base64:) regex stack srfi-13 irregex) (import commonmod) (import dbmod) +(import servermod) ;; (use (prefix ulex ulex:)) ;; These are called by the server on recipt of /api calls ;; - keep it simple, only return the actual result of the call, i.e. no meta info here Index: commonmod.scm ================================================================== --- commonmod.scm +++ commonmod.scm @@ -39,20 +39,40 @@ (prefix mtconfigf configf:) ;; stml2 ;; (prefix margs args:) z3 (prefix base64 base64:) (prefix mtargs args:) - directory-utils) + directory-utils + sparse-vectors) (include "common_records.scm") (include "megatest-fossil-hash.scm") (include "megatest-version.scm") ;; no need to export this (define *verbosity-cache* (make-hash-table)) (define *verbosity* 0) + +(defstruct remote + (hh-dat (common:get-homehost)) ;; homehost record ( addr . hhflag ) + (server-url #f) ;; (if *toppath* (server:check-if-running *toppath*))) ;; (server:check-if-running *toppath*) #f)) + (last-server-check 0) ;; last time we checked to see if the server was alive + (conndat #f) + (transport *transport-type*) + (server-timeout #f) ;; (server:expiration-timeout)) + (force-server #f) + (ro-mode #f) + (ro-mode-checked #f)) ;; flag that indicates we have checked for ro-mode + +;; launching and hosts +(defstruct host + (reachable #f) + (last-update 0) + (last-used 0) + (last-cpuload 1)) + ;; GLOBALS ;; CONTEXTS @@ -1666,29 +1686,10 @@ fullpath)) (define *common:this-exe-fullpath* (common:get-this-exe-fullpath)) (define *common:this-exe-dir* (pathname-directory *common:this-exe-fullpath*)) (define *common:this-exe-name* (pathname-strip-directory *common:this-exe-fullpath*)) -;; Move me elsewhere ... -;; RADT => Why do we meed the version check here, this is called only if version misma -;; -(define (common:cleanup-db dbstruct #!key (full #f)) - (apply db:multi-db-sync - dbstruct - 'schema - ;; 'new2old - 'killservers - 'adj-target - ;; 'old2new - 'new2old - ;; (if full - '(dejunk) - ;; '()) - ) - (if (common:api-changed?) - (common:set-last-run-version))) - (define (common:snapshot-file filepath #!key (subdir ".") ) (if (file-exists? filepath) (let* ((age-sec (lambda (file) (if (file-exists? file) (- (current-seconds) (file-modification-time file)) @@ -1848,56 +1849,10 @@ (debug:print-error 0 *default-log-port* "failed to remove " fullname) (delete-file* fullname))))) files) (debug:print-info 0 *default-log-port* "Deleted " (length files) " files from logs, keeping " max-allowed " files.")))))) -;; Force a megatest cleanup-db if version is changed and skip-version-check not specified -;; Do NOT check if not on homehost! -;; -(define (common:exit-on-version-changed) - (if (common:on-homehost?) - (if (common:api-changed?) - (let* ((mtconf (conc (get-environment-variable "MT_RUN_AREA_HOME") "/megatest.config")) - (dbfile (conc (get-environment-variable "MT_RUN_AREA_HOME") "/megatest.db")) - (read-only (not (file-write-access? dbfile))) - (dbstruct (db:setup #t))) - (debug:print 0 *default-log-port* - "WARNING: Version mismatch!\n" - " expected: " (common:version-signature) "\n" - " got: " (common:get-last-run-version)) - (cond - ((get-environment-variable "MT_SKIP_DB_MIGRATE") #t) - ((and (common:file-exists? mtconf) (common:file-exists? dbfile) (not read-only) - (eq? (current-user-id)(file-owner mtconf))) ;; safe to run -cleanup-db - (debug:print 0 *default-log-port* " I see you are the owner of megatest.config, attempting to cleanup and reset to new version") - (handle-exceptions - exn - (begin - (debug:print 0 *default-log-port* "Failed to switch versions.") - (debug:print 0 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn)) - (print-call-chain (current-error-port)) - (exit 1)) - (common:cleanup-db dbstruct))) - ((not (common:file-exists? mtconf)) - (debug:print 0 *default-log-port* " megatest.config does not exist in this area. Cannot proceed with megatest version migration.") - (exit 1)) - ((not (common:file-exists? dbfile)) - (debug:print 0 *default-log-port* " megatest.db does not exist in this area. Cannot proceed with megatest version migration.") - (exit 1)) - ((not (eq? (current-user-id)(file-owner mtconf))) - (debug:print 0 *default-log-port* " You do not own megatest.db in this area. Cannot proceed with megatest version migration.") - (exit 1)) - (read-only - (debug:print 0 *default-log-port* " You have read-only access to this area. Cannot proceed with megatest version migration.") - (exit 1)) - (else - (debug:print 0 *default-log-port* " to switch versions you can run: \"megatest -cleanup-db\"") - (exit 1))))))) -;; (begin -;; (debug:print 0 *default-log-port* "ERROR: cannot migrate version unless on homehost. Exiting.") -;; (exit 1)))) - ;;====================================================================== ;; S P A R S E A R R A Y S ;;====================================================================== (define (make-sparse-array) @@ -1983,202 +1938,10 @@ (handle-exceptions exn #f ;; I don't really care why this failed (at least for now) (delete-file* fname))) -;;====================================================================== -;; D E B U G G I N G S T U F F -;;====================================================================== - -;; (define *verbosity* 1) -;; (define *logging* #f) - -(define (common:set-last-run-version) - (rmt:set-var "MEGATEST_VERSION" (common:version-signature))) - -;; postive number if megatest version > db version -;; negative number if megatest version < db version -(define (common:version-db-delta) - (- megatest-version (common:get-last-run-version-number))) - -(define (common:version-changed?) - (not (equal? (common:get-last-run-version) - (common:version-signature)))) - -;; from metadat lookup MEGATEST_VERSION -;; -(define (common:get-last-run-version) ;; RADT => How does this work in send-receive function??; assume it is the value saved in some DB - (rmt:get-var "MEGATEST_VERSION")) - -(define (common:get-last-run-version-number) - (string->number - (substring (common:get-last-run-version) 0 6))) - -(define (common:api-changed?) - (not (equal? (substring (->string megatest-version) 0 4) - (substring (conc (common:get-last-run-version)) 0 4)))) - - -;;====================================================================== -;; E X I T H A N D L I N G -;;====================================================================== - -(define (common:run-sync?) - (and (common:on-homehost?) - (args:get-arg "-server"))) - -;; (let ((ohh (common:on-homehost?)) -;; (srv (args:get-arg "-server"))) -;; (and ohh srv))) - ;; (debug:print-info 0 *default-log-port* "common:run-sync? ohh=" ohh ", srv=" srv) - -(define *watchdog* (make-thread - (lambda () - (handle-exceptions - exn - (begin - (print-call-chain) - (print " message: " ((condition-property-accessor 'exn 'message) exn))) - (common:watchdog))) - "Watchdog thread")) - -;; currently the primary job of the watchdog is to run the sync back to megatest.db from the db in /tmp -;; if we are on the homehost and we are a server (by definition we are on the homehost if we are a server) -;; -(define (common:readonly-watchdog dbstruct) - (thread-sleep! 0.05) ;; delay for startup - (debug:print-info 13 *default-log-port* "common:readonly-watchdog entered.") - ;; sync megatest.db to /tmp/.../megatst.db - (let* ((sync-cool-off-duration 3) - (golden-mtdb (dbr:dbstruct-mtdb dbstruct)) - (golden-mtpath (db:dbdat-get-path golden-mtdb)) - (tmp-mtdb (dbr:dbstruct-tmpdb dbstruct)) - (tmp-mtpath (db:dbdat-get-path tmp-mtdb))) - (debug:print-info 0 *default-log-port* "Read-only periodic sync thread started.") - (let loop ((last-sync-time 0)) - (debug:print-info 13 *default-log-port* "loop top tmp-mtpath="tmp-mtpath" golden-mtpath="golden-mtpath) - (let* ((duration-since-last-sync (- (current-seconds) last-sync-time))) - (debug:print-info 13 *default-log-port* "duration-since-last-sync="duration-since-last-sync) - (if (and (not *time-to-exit*) - (< duration-since-last-sync sync-cool-off-duration)) - (thread-sleep! (- sync-cool-off-duration duration-since-last-sync))) - (if (not *time-to-exit*) - (let ((golden-mtdb-mtime (file-modification-time golden-mtpath)) - (tmp-mtdb-mtime (file-modification-time tmp-mtpath))) - (if (> golden-mtdb-mtime tmp-mtdb-mtime) - (if (< golden-mtdb-mtime (- (current-seconds) 3)) ;; file has NOT been touched in past three seconds, this way multiple servers won't fight to sync back - (let ((res (db:multi-db-sync dbstruct 'old2new))) - (debug:print-info 13 *default-log-port* "rosync called, " res " records transferred.")))) - (loop (current-seconds))) - #t))) - (debug:print-info 0 *default-log-port* "Exiting readonly-watchdog timer, *time-to-exit* = " *time-to-exit*" pid="(current-process-id)" mtpath="golden-mtpath))) - -;; TODO: for multiple areas, we will have multiple watchdogs; and multiple threads to manage -(define (common:watchdog) - (debug:print-info 13 *default-log-port* "common:watchdog entered.") - (if (launch:setup) - (if (common:on-homehost?) - (let ((dbstruct (db:setup #t))) - (debug:print-info 13 *default-log-port* "after db:setup with dbstruct=" dbstruct) - (cond - ((dbr:dbstruct-read-only dbstruct) - (debug:print-info 13 *default-log-port* "loading read-only watchdog") - (common:readonly-watchdog dbstruct)) - (else - (debug:print-info 13 *default-log-port* "loading writable-watchdog.") - (let* ((syncer (or (configf:lookup *configdat* "server" "sync-method") "brute-force-sync"))) - (cond - ((equal? syncer "brute-force-sync") - (server:writable-watchdog-bruteforce dbstruct)) - ((equal? syncer "delta-sync") - (server:writable-watchdog-deltasync dbstruct)) - (else - (debug:print-error 0 *default-log-port* "Unknown server/sync-method specified ("syncer") - valid values are brute-force-sync and delta-sync.") - (exit 1))) - ;;(debug:print 1 *default-log-port* "INFO: ["(common:human-time)"] Syncer started (method="syncer")") - ))) - (debug:print-info 13 *default-log-port* "watchdog done.")) - (debug:print-info 13 *default-log-port* "no need for watchdog on non-homehost")))) - - -(define (std-exit-procedure) - ;;(common:telemetry-log-close) - (on-exit (lambda () 0)) - ;;(debug:print-info 13 *default-log-port* "std-exit-procedure called; *time-to-exit*="*time-to-exit*) - (let ((no-hurry (if *time-to-exit* ;; hurry up - #f - (begin - (set! *time-to-exit* #t) - #t)))) - (debug:print-info 4 *default-log-port* "starting exit process, finalizing databases.") - (if (and no-hurry (debug:debug-mode 18)) - (rmt:print-db-stats)) - (let ((th1 (make-thread (lambda () ;; thread for cleaning up, give it five seconds - (if *dbstruct-db* (db:close-all *dbstruct-db*)) ;; one second allocated - (if *task-db* - (let ((db (cdr *task-db*))) - (if (sqlite3:database? db) - (begin - (sqlite3:interrupt! db) - (sqlite3:finalize! db #t) - ;; (vector-set! *task-db* 0 #f) - (set! *task-db* #f))))) - (http-client#close-all-connections!) - ;; (if (and *runremote* - ;; (remote-conndat *runremote*)) - ;; (begin - ;; (http-client#close-all-connections!))) ;; for http-client - (if (not (eq? *default-log-port* (current-error-port))) - (close-output-port *default-log-port*)) - (set! *default-log-port* (current-error-port))) "Cleanup db exit thread")) - (th2 (make-thread (lambda () - (debug:print 4 *default-log-port* "Attempting clean exit. Please be patient and wait a few seconds...") - (if no-hurry - (begin - (thread-sleep! 5)) ;; give the clean up few seconds to do it's stuff - (begin - (thread-sleep! 2))) - (debug:print 4 *default-log-port* " ... done") - ) - "clean exit"))) - (thread-start! th1) - (thread-start! th2) - (thread-join! th1) - ) - ) - - 0) - -(define (std-signal-handler signum) - ;; (signal-mask! signum) - (set! *time-to-exit* #t) - ;;(debug:print-info 13 *default-log-port* "got signal "signum) - (debug:print-error 0 *default-log-port* "Received signal " signum " aaa exiting promptly") - ;; (std-exit-procedure) ;; shouldn't need this since we are exiting and it will be called anyway - (exit)) - -(define (special-signal-handler signum) - ;; (signal-mask! signum) - (set! *time-to-exit* #t) - ;;(debug:print-info 13 *default-log-port* "got signal "signum) - (debug:print-error 0 *default-log-port* "Received signal " signum " sending email befor exiting!!") - ;;TODO send email to notify admin contact listed in the config that the lisner got killed - ;; (std-exit-procedure) ;; shouldn't need this since we are exiting and it will be called anyway - (exit)) - - -(set-signal-handler! signal/int std-signal-handler) ;; ^C -(set-signal-handler! signal/term std-signal-handler) - -;; (set-signal-handler! signal/stop std-signal-handler) ;; ^Z NO, do NOT handle ^Z! - -;; '(print (string-intersperse (map cadr (hash-table-ref/default (read-config "megatest.config" \#f \#t) "disks" '"'"'("none" ""))) "\n"))' -(define (common:get-disks #!key (configf #f)) - (hash-table-ref/default - (or configf (read-config "megatest.config" #f #t)) - "disks" '("none" ""))) - ;;====================================================================== ;; T A R G E T S , S T A T E , S T A T U S , ;; R U N N A M E A N D T E S T P A T T ;;====================================================================== @@ -2185,11 +1948,11 @@ ;; (map print (map car (hash-table->alist (read-config "runconfigs.config" #f #t)))) ;; (define (common:get-runconfig-targets #!key (configf #f)) (let ((targs (sort (map car (hash-table->alist (or configf ;; NOTE: There is no value in using runconfig:read here. - (read-config (conc *toppath* "/runconfigs.config") + (configf:read-config (conc *toppath* "/runconfigs.config") #f #t) (make-hash-table)))) string trynum 0) - (begin - (thread-sleep! 2) - (common:get-homehost trynum: (- trynum 1))) - #f)) + (debug:print-error 0 *default-log-port* "common:get-homehost called but *toppath* not defined!! Exiting.") + (exit)) +;; (mutex-unlock! *homehost-mutex*) +;; (launch:setup) ;; safely mutexed now +;; (if (> trynum 0) +;; (begin +;; (thread-sleep! 2) +;; (common:get-homehost trynum: (- trynum 1))) +;; #f)) (else (let* ((currhost (get-host-name)) (bestadrs (server:get-best-guess-address currhost)) ;; first look in config, then look in file .homehost, create it if not found (homehost (or (configf:lookup *configdat* "server" "homehost" ) @@ -2567,193 +2332,10 @@ (define (common:unix-ping hostname) (let ((res (system (conc "ping -c 1 " hostname " > /dev/null")))) (eq? res 0))) -;; ideally put all this info into the db, no need to preserve it across moving homehost -;; -;; return list of -;; ( reachable? cpuload update-time ) -(define (common:get-host-info hostname) - (let* ((loadinfo (rmt:get-latest-host-load hostname)) ;; if this host happens to have been recently used by a test reuse the load data - (load (car loadinfo)) - (load-sample-time (cdr loadinfo)) - (load-sample-age (- (current-seconds) load-sample-time)) - (loadinfo-timeout-seconds 6) ;; this was 20 seconds, seems way too lax. Switch to 6 seconds - (host-last-update-timeout-seconds 4) - (host-rec (hash-table-ref/default *host-loads* hostname #f)) - ) - (cond - ((< load-sample-age loadinfo-timeout-seconds) - (list #t - load-sample-time - load)) - ((and host-rec - (< (current-seconds) (+ (host-last-update host-rec) host-last-update-timeout-seconds))) - (list #t - (host-last-update host-rec) - (host-last-cpuload host-rec ))) - ((common:unix-ping hostname) - (list #t - (current-seconds) - (alist-ref 'adj-core-load (common:get-normalized-cpu-load hostname)))) ;; this is cheaper than you might think. get-normalized-cpu-load is cached for up to 5 seconds - (else - (list #f 0 -1) ;; bad host, don't use! - )))) - -;; see defstruct host at top of file. -;; host: reachable last-update last-used last-cpuload -;; -(define (common:update-host-loads-table hosts-raw) - (let* ((hosts (filter (lambda (x) - (string-match (regexp "^\\S+$") x)) - hosts-raw))) - (for-each - (lambda (hostname) - (let* ((rec (let ((h (hash-table-ref/default *host-loads* hostname #f))) - (if h - h - (let ((h (make-host))) - (hash-table-set! *host-loads* hostname h) - h)))) - (host-info (common:get-host-info hostname)) - (is-reachable (car host-info)) - (last-reached-time (cadr host-info)) - (load (caddr host-info))) - (host-reachable-set! rec is-reachable) - (host-last-update-set! rec last-reached-time) - (host-last-cpuload-set! rec load))) - hosts))) - -;; go through the hosts from least recently used to most recently used, pick the first that meets the load criteral from the -;; [host-rules] section. -;; -(define (common:get-least-loaded-host hosts-raw host-type configdat) - (let* ((rdat (configf:lookup configdat "host-rules" host-type)) - (rules (common:val->alist (or rdat "") convert: #t)) ;; maxnload, maxnjobs, maxjobrate - (maxnload (common:alist-ref/default 'maxnload rules 1.5)) ;; max normalized load - (maxnjobs (common:alist-ref/default 'maxnjobs rules 1.5)) ;; max normalized number of jobs - (maxjobrate (common:alist-ref/default 'maxjobrate rules (/ 1 6))) ;; max rate of submitting jobs to a given host in jobs/second - (hosts (filter (lambda (x) - (string-match (regexp "^\\S+$") x)) - hosts-raw)) - ;; (best-host #f) - (get-rec (lambda (hostname) - ;; (print "get-rec hostname=" hostname) - (let ((h (hash-table-ref/default *host-loads* hostname #f))) - (if h - h - (let ((h (make-host))) - (hash-table-set! *host-loads* hostname h) - h))))) - (best-load 99999) - (curr-time (current-seconds)) - (get-hosts-sorted (lambda (hosts) - (sort hosts (lambda (a b) - (let ((a-rec (get-rec a)) - (b-rec (get-rec b))) - ;; (print "a=" a " a-rec=" a-rec " host-last-used=" (host-last-used a-rec)) - ;; (print "b=" b " b-rec=" b-rec " host-last-used=" (host-last-used b-rec)) - (< (host-last-used a-rec) - (host-last-used b-rec)))))))) - (debug:print 0 *default-log-port* "INFO: hosts-sorted=" (get-hosts-sorted hosts)) - (if (null? hosts) - #f ;; no hosts to select from. All done and giving up now. - (let ((hosts-sorted (get-hosts-sorted hosts))) - (common:update-host-loads-table hosts) - (let loop ((hostname (car hosts-sorted)) - (tal (cdr hosts-sorted)) - (best-host #f)) - (let* ((rec (get-rec hostname)) - (reachable (host-reachable rec)) - (load (host-last-cpuload rec)) - (last-used (host-last-used rec)) - (delta (- curr-time last-used)) - (job-rate (if (> delta 0) - (/ 1 delta) - 999)) ;; jobs per second - (new-best - (cond - ((not reachable) - (debug:print 0 *default-log-port* "Skipping host " hostname " as it cannot be reached.") - best-host) - ((and (< load maxnload) ;; load is acceptable - (< job-rate maxjobrate)) ;; job rate is acceptable - (set! best-load load) - hostname) - (else best-host)))) - (debug:print 0 *default-log-port* "INFO: Trying host " hostname " with load " load ", last used " delta " seconds ago, with job-rate " job-rate " for running a test." ) - (if new-best - (begin ;; found a host, return it - (debug:print 0 *default-log-port* "INFO: Found host: " new-best " load: " load " last-used: " delta " seconds ago, with job-rate: " job-rate) - (host-last-used-set! rec curr-time) - new-best) - (if (null? tal) #f (loop (car tal)(cdr tal) best-host))))))))) - -(define (common:wait-for-cpuload maxload-in numcpus-in waitdelay #!key (count 1000) (msg #f)(remote-host #f)(force-maxload #f)) - (let* ((loadavg (common:get-cpu-load remote-host)) - (numcpus (if (<= 1 numcpus-in) ;; not possible to have zero. If we get 1, it's possible that we got the previous default, and we should check again - (common:get-num-cpus remote-host) - numcpus-in)) - (maxload (if force-maxload - maxload-in - (max maxload-in 0.5))) ;; so maxload must be greater than 0.5 for now BUG - FIXME? - (first (car loadavg)) - (next (cadr loadavg)) - (adjload (* maxload (max 1 numcpus))) ;; possible bug where numcpus (or could be maxload) is zero, crude fallback is to at least use 1 - (loadjmp (- first next)) - (adjwait (min (+ 300 (random 10)) (abs (* (+ (random 10)(/ (- 1000 count) 10) waitdelay) (- first adjload) )) ))) ;; add some randomness to the time to break any alignment where netbatch dumps many jobs to machines simultaneously - (debug:print-info 1 *default-log-port* "Checking cpuload on " (or remote-host "localhost") ", maxload: " maxload - ", load: " first ", adjload: " adjload ", loadjmp: " loadjmp) - (cond - ((and (> first adjload) - (> count 0)) - (debug:print-info 0 *default-log-port* "server start delayed " adjwait " seconds due to load " first " exceeding max of " adjload " on server " (or remote-host (get-host-name)) " (normalized load-limit: " maxload ") " (if msg msg "")) - (thread-sleep! adjwait) - (common:wait-for-cpuload maxload numcpus waitdelay count: (- count 1) msg: msg remote-host: remote-host)) - ((and (> loadjmp numcpus) - (> count 0)) - (debug:print-info 0 *default-log-port* "waiting " adjwait " seconds due to load jump " loadjmp " > numcpus " numcpus (if msg msg "")) - (thread-sleep! adjwait) - (common:wait-for-cpuload maxload numcpus waitdelay count: (- count 1) msg: msg remote-host: remote-host))))) - -(define (common:wait-for-homehost-load maxload msg) - (let* ((hh-dat (if (common:on-homehost?) ;; if we are on the homehost then pass in #f so the calls are local. - #f - (common:get-homehost))) - (hh (if hh-dat (car hh-dat) #f)) - (numcpus (common:get-num-cpus hh))) - (common:wait-for-normalized-load maxload msg hh))) - -(define (common:get-num-cpus remote-host) - (let* ((actual-host (or remote-host (get-host-name)))) - (or (common:get-cached-info actual-host "num-cpus" age: 86400) ;; hosts had better not be changing the number of cpus too often! - (let* ((proc (lambda () - (let loop ((numcpu 0) - (inl (read-line))) - (if (eof-object? inl) - (begin - (common:write-cached-info remote-host "num-cpus" numcpu) - numcpu) - (loop (if (string-match "^processor\\s+:\\s+\\d+$" inl) - (+ numcpu 1) - numcpu) - (read-line)))))) - (result (if remote-host - (with-input-from-pipe - (conc "ssh " remote-host " cat /proc/cpuinfo") - proc) - (with-input-from-file "/proc/cpuinfo" proc)))) - (common:write-cached-info actual-host "num-cpus" result) - result)))) - -;; wait for normalized cpu load to drop below maxload -;; -(define (common:wait-for-normalized-load maxload msg remote-host) - (let ((num-cpus (common:get-num-cpus remote-host))) - (common:wait-for-cpuload maxload num-cpus 15 msg: msg remote-host: remote-host))) - (define (get-uname . params) (let* ((uname-res (process:cmd-run->list (conc "uname " (if (null? params) "-a" (car params))))) (uname #f)) (if (null? (car uname-res)) "unknown" @@ -2777,10 +2359,23 @@ ;; ;; (process-wait pid) ;; res) ;; (loop (read-line) inl)))))) (with-input-from-pipe (conc "readlink -f " inpath) read-line)) +(define (server:get-best-guess-address hostname) + (let ((res #f)) + (for-each + (lambda (adr) + (if (not (eq? (u8vector-ref adr 0) 127)) + (set! res adr))) + ;; NOTE: This can fail when there is no mention of the host in /etc/hosts. FIXME + (vector->list (hostinfo-addresses (hostname->hostinfo hostname)))) + (string-intersperse + (map number->string + (u8vector->list + (if res res (hostname->ip hostname)))) "."))) + ;;====================================================================== ;; D I S K S P A C E ;;====================================================================== (define (common:get-disk-space-used fpath) @@ -3313,14 +2908,14 @@ (define (common:load-views-config) (let* ((view-cfgdat (make-hash-table)) (home-cfgfile (conc (get-environment-variable "HOME") "/.mtviews.config")) (mthome-cfgfile (conc *toppath* "/.mtviews.config"))) (if (common:file-exists? mthome-cfgfile) - (read-config mthome-cfgfile view-cfgdat #t)) + (configf:read-config mthome-cfgfile view-cfgdat #t)) ;; we load the home dir file AFTER the MTRAH file so the user can clobber settings when running the dashboard in read-only areas (if (common:file-exists? home-cfgfile) - (read-config home-cfgfile view-cfgdat #t)) + (configf:read-config home-cfgfile view-cfgdat #t)) view-cfgdat)) ;;====================================================================== ;; H I E R A R C H I C A L H A S H T A B L E S ;;====================================================================== Index: dbmod.scm ================================================================== --- dbmod.scm +++ dbmod.scm @@ -24,10 +24,11 @@ (declare (uses tasksmod)) (declare (uses odsmod)) (declare (uses testsmod)) (declare (uses mtargs)) (declare (uses mtconfigf)) +;; (declare (uses servermod)) (module dbmod * (import scheme chicken data-structures extras) @@ -46,10 +47,11 @@ (import tasksmod) (import odsmod) (import testsmod) (import (prefix mtargs args:)) (import (prefix mtconfigf configf:)) +;; (import servermod) ;; (use (prefix ulex ulex:)) (include "common_records.scm") (include "db_records.scm") @@ -1004,130 +1006,10 @@ ;; use-last-update: #t))) ;; (thread-start! th1) ;; (apply proc cache-db params) ;; )))) -;; options: -;; -;; 'killservers - kills all servers -;; 'dejunk - removes junk records -;; 'adj-testids - move test-ids into correct ranges -;; 'old2new - sync megatest.db to /tmp/.../megatest.db and /tmp/.../megatest_ref.db -;; 'new2old - sync /tmp/.../megatest.db to megatest.db and /tmp/.../megatest_ref.db (and update data_synced) -;; 'closeall - close all opened dbs -;; 'schema - attempt to apply schema changes -;; run-ids: '(1 2 3 ...) or #f (for all) -;; -(define (db:multi-db-sync dbstruct . options) - ;; (if (not (launch:setup)) - ;; (debug:print 0 *default-log-port* "ERROR: not able to setup up for megatest.") - (let* ((mtdb (dbr:dbstruct-mtdb dbstruct)) - (tmpdb (db:get-db dbstruct)) - (refndb (dbr:dbstruct-refndb dbstruct)) - (allow-cleanup #t) ;; (if run-ids #f #t)) - (servers (server:get-list *toppath*)) ;; (tasks:get-all-servers (db:delay-if-busy tdbdat))) - (data-synced 0)) ;; count of changed records (I hope) - - (for-each - (lambda (option) - - (case option - ;; kill servers - ((killservers) - (for-each - (lambda (server) - (match-let (((mod-time host port start-time pid) server)) - (if (and host pid) - (tasks:kill-server host pid)))) - servers) - - ;; /tmp/bjbarcla/megatest_localdb/fullrun/.nfs.pdx.disks.icf_env_disk001.bjbarcla.gwa.issues.mtdev.matt-bisect.megatest.ext-tests.runs.all.v1.65.1.6524.dev.bb-24-justrollup-f8.rollup.fullrun/megatest.db.lock - (delete-file* (common:get-sync-lock-filepath)) - ) - - ;; clear out junk records - ;; - ((dejunk) - (db:delay-if-busy mtdb) ;; ok to delay on mtdb - (when (file-write-access? (db:dbdat-get-path mtdb)) (db:clean-up mtdb)) - (db:clean-up tmpdb) - (db:clean-up refndb)) - - ;; sync runs, test_meta etc. - ;; - ((old2new) - (set! data-synced - (+ (db:sync-tables (db:sync-all-tables-list dbstruct) #f mtdb tmpdb refndb) - data-synced))) - - ;; now ensure all newdb data are synced to megatest.db - ;; do not use the run-ids list passed in to the function - ;; - ((new2old) - (set! data-synced - (+ (db:sync-tables (db:sync-all-tables-list dbstruct) #f tmpdb refndb mtdb) - data-synced))) - - ((adj-target) - (db:adj-target (db:dbdat-get-db mtdb)) - (db:adj-target (db:dbdat-get-db tmpdb)) - (db:adj-target (db:dbdat-get-db refndb))) - - ((schema) - (db:patch-schema-maindb (db:dbdat-get-db mtdb)) - (db:patch-schema-maindb (db:dbdat-get-db tmpdb)) - (db:patch-schema-maindb (db:dbdat-get-db refndb)) - (db:patch-schema-rundb (db:dbdat-get-db mtdb)) - (db:patch-schema-rundb (db:dbdat-get-db tmpdb)) - (db:patch-schema-rundb (db:dbdat-get-db refndb)))) - - (stack-push! (dbr:dbstruct-dbstack dbstruct) tmpdb)) - options) - data-synced)) - -(define (db:tmp->megatest.db-sync dbstruct last-update) - (let* ((mtdb (dbr:dbstruct-mtdb dbstruct)) - (tmpdb (db:get-db dbstruct)) - (refndb (dbr:dbstruct-refndb dbstruct)) - (res (db:sync-tables (db:sync-all-tables-list dbstruct) last-update tmpdb refndb mtdb))) - (stack-push! (dbr:dbstruct-dbstack dbstruct) tmpdb) - res)) - -;;;; run-ids -;; if #f use *db-local-sync* : or 'local-sync-flags -;; if #t use timestamps : or 'timestamps -;; -;; NB// no-sync-db is the db handle, not a flag! -;; -(define (db:sync-to-megatest.db dbstruct #!key (no-sync-db #f)) - (let* ((start-time (current-seconds)) - (last-full-update (if no-sync-db - (db:no-sync-get/default no-sync-db "LAST_FULL_UPDATE" 0) - 0)) - (full-sync-needed (> (- start-time last-full-update) 3600)) ;; every hour do a full sync - (last-update (if full-sync-needed - 0 - (if no-sync-db - (db:no-sync-get/default no-sync-db "LAST_UPDATE" 0) - 0))) ;; (or (db:get-var dbstruct "LAST_UPDATE") 0)) - (sync-needed (> (- start-time last-update) 6)) - (res (if (or sync-needed ;; don't sync if a sync already occurred in the past 6 seconds - full-sync-needed) - (begin - (if no-sync-db - (begin - (if full-sync-needed (db:no-sync-set no-sync-db "LAST_FULL_UPDATE" start-time)) - (db:no-sync-set no-sync-db "LAST_UPDATE" start-time))) - (db:tmp->megatest.db-sync dbstruct last-update)) - 0)) - (sync-time (- (current-seconds) start-time))) - (debug:print-info 3 *default-log-port* "Sync of newdb to olddb completed in " sync-time " seconds pid="(current-process-id)) - (if (common:low-noise-print 30 "sync new to old") - (if sync-needed - (debug:print-info 0 *default-log-port* "Sync of " res " records from newdb to olddb completed in " sync-time " seconds pid="(current-process-id)) - (debug:print-info 0 *default-log-port* "No sync needed, last updated " (- start-time last-update) " seconds ago"))) - res)) ;; keeping it around for debugging purposes only #;(define (open-run-close-no-exception-handling proc idb . params) (debug:print-info 11 *default-log-port* "open-run-close-no-exception-handling START given a db=" (if idb "yes " "no ") ", params=" params) (print "I don't work anymore. open-run-close-no-exception-handling needs fixing or removing...") Index: docs/code/module-hierarchy.dot ================================================================== --- docs/code/module-hierarchy.dot +++ docs/code/module-hierarchy.dot @@ -31,11 +31,12 @@ mtconfigf -> commonmod; processmod -> commonmod; stml2 -> commonmod; // apimod commonmod -> apimod; - dbmod -> apimod; + dbmod -> apimod; + servermod -> apimod; // rmtmod dbmod -> rmtmod; commonmod -> rmtmod; apimod -> rmtmod; ulex -> rmtmod; @@ -68,12 +69,13 @@ // pgdbmod commonmod -> pgdbmod; mtconfigf -> pgdbmod; //servermod commonmod -> servermod; + dbmod -> servermod; // processmod } } Index: mtmod.scm ================================================================== --- mtmod.scm +++ mtmod.scm @@ -669,9 +669,529 @@ (rmt:no-sync-get-lock keyname)) (define (common:simple-unlock keyname #!key (force #f)) (rmt:no-sync-del! keyname)) +;;====================================================================== +;; db based host calls +;;====================================================================== + +;;====================================================================== +;; T E S T L A U N C H I N G P E R I T E M W I T H H O S T T Y P E S +;;====================================================================== +;; +;; [hosts] +;; arm cubie01 cubie02 +;; x86_64 zeus xena myth01 +;; allhosts #{g hosts arm} #{g hosts x86_64} +;; +;; [host-types] +;; general #MTLOWESTLOAD #{g hosts allhosts} +;; arm #MTLOWESTLOAD #{g hosts arm} +;; nbgeneral nbjob run JOBCOMMAND -log $MT_LINKTREE/$MT_TARGET/$MT_RUNNAME.$MT_TESTNAME-$MT_ITEM_PATH.lgo +;; +;; [host-rules] +;; # maxnload => max normalized load +;; # maxnjobs => max jobs per cpu +;; # maxjobrate => max jobs per second +;; general maxnload=1.1; maxnjobs=1.2; maxjobrate=0.1 +;; +;; [launchers] +;; envsetup general +;; xor/%/n 4C16G +;; % nbgeneral +;; +;; [jobtools] +;; # if defined and not "no" flexi-launcher will bypass "launcher" unless no match. +;; flexi-launcher yes +;; launcher nbfake +;; +(define (common:get-launcher configdat testname itempath) + (let ((fallback-launcher (configf:lookup configdat "jobtools" "launcher"))) + (if (and (configf:lookup configdat "jobtools" "flexi-launcher") ;; overrides launcher + (not (equal? (configf:lookup configdat "jobtools" "flexi-launcher") "no"))) + (let* ((launchers (hash-table-ref/default configdat "launchers" '()))) + (if (null? launchers) + fallback-launcher + (let loop ((hed (car launchers)) + (tal (cdr launchers))) + (let ((patt (car hed)) + (host-type (cadr hed))) + (if (tests:match patt testname itempath) + (begin + (debug:print-info 2 *default-log-port* "Have flexi-launcher match for " testname "/" itempath " = " host-type) + (let ((launcher (configf:lookup configdat "host-types" host-type))) + (if launcher + (let* ((launcher-parts (string-split launcher)) + (launcher-exe (car launcher-parts))) + (if (equal? launcher-exe "#MTLOWESTLOAD") ;; this is our special case, we will find the lowest load and craft a nbfake commandline + (let host-loop ((targ-host (common:get-least-loaded-host (cdr launcher-parts) host-type configdat)) + (count 100)) + (if targ-host + (conc "remrun " targ-host) + (if (> count 0) + (begin + (debug:print 0 *default-log-port* "INFO: Waiting for a host for host-type " host-type) + (thread-sleep! (- 101 count)) + (host-loop (common:get-least-loaded-host (cdr launcher-parts) host-type configdat) + (- count 1))) + (begin + (debug:print 0 *default-log-port* "FATAL: Failed to find a host from #MTLOWESTLOAD for host-type " host-type) + (exit))))) + launcher)) + (begin + (debug:print-info 0 *default-log-port* "WARNING: no launcher found for host-type " host-type) + (if (null? tal) + fallback-launcher + (loop (car tal)(cdr tal))))))) + ;; no match, try again + (if (null? tal) + fallback-launcher + (loop (car tal)(cdr tal)))))))) + fallback-launcher))) + +;; ideally put all this info into the db, no need to preserve it across moving homehost +;; +;; return list of +;; ( reachable? cpuload update-time ) +(define (common:get-host-info hostname) + (let* ((loadinfo (rmt:get-latest-host-load hostname)) ;; if this host happens to have been recently used by a test reuse the load data + (load (car loadinfo)) + (load-sample-time (cdr loadinfo)) + (load-sample-age (- (current-seconds) load-sample-time)) + (loadinfo-timeout-seconds 6) ;; this was 20 seconds, seems way too lax. Switch to 6 seconds + (host-last-update-timeout-seconds 4) + (host-rec (hash-table-ref/default *host-loads* hostname #f)) + ) + (cond + ((< load-sample-age loadinfo-timeout-seconds) + (list #t + load-sample-time + load)) + ((and host-rec + (< (current-seconds) (+ (host-last-update host-rec) host-last-update-timeout-seconds))) + (list #t + (host-last-update host-rec) + (host-last-cpuload host-rec ))) + ((common:unix-ping hostname) + (list #t + (current-seconds) + (alist-ref 'adj-core-load (common:get-normalized-cpu-load hostname)))) ;; this is cheaper than you might think. get-normalized-cpu-load is cached for up to 5 seconds + (else + (list #f 0 -1) ;; bad host, don't use! + )))) + +;; see defstruct host at top of file. +;; host: reachable last-update last-used last-cpuload +;; +(define (common:update-host-loads-table hosts-raw) + (let* ((hosts (filter (lambda (x) + (string-match (regexp "^\\S+$") x)) + hosts-raw))) + (for-each + (lambda (hostname) + (let* ((rec (let ((h (hash-table-ref/default *host-loads* hostname #f))) + (if h + h + (let ((h (make-host))) + (hash-table-set! *host-loads* hostname h) + h)))) + (host-info (common:get-host-info hostname)) + (is-reachable (car host-info)) + (last-reached-time (cadr host-info)) + (load (caddr host-info))) + (host-reachable-set! rec is-reachable) + (host-last-update-set! rec last-reached-time) + (host-last-cpuload-set! rec load))) + hosts))) + +;; go through the hosts from least recently used to most recently used, pick the first that meets the load criteral from the +;; [host-rules] section. +;; +(define (common:get-least-loaded-host hosts-raw host-type configdat) + (let* ((rdat (configf:lookup configdat "host-rules" host-type)) + (rules (common:val->alist (or rdat "") convert: #t)) ;; maxnload, maxnjobs, maxjobrate + (maxnload (common:alist-ref/default 'maxnload rules 1.5)) ;; max normalized load + (maxnjobs (common:alist-ref/default 'maxnjobs rules 1.5)) ;; max normalized number of jobs + (maxjobrate (common:alist-ref/default 'maxjobrate rules (/ 1 6))) ;; max rate of submitting jobs to a given host in jobs/second + (hosts (filter (lambda (x) + (string-match (regexp "^\\S+$") x)) + hosts-raw)) + ;; (best-host #f) + (get-rec (lambda (hostname) + ;; (print "get-rec hostname=" hostname) + (let ((h (hash-table-ref/default *host-loads* hostname #f))) + (if h + h + (let ((h (make-host))) + (hash-table-set! *host-loads* hostname h) + h))))) + (best-load 99999) + (curr-time (current-seconds)) + (get-hosts-sorted (lambda (hosts) + (sort hosts (lambda (a b) + (let ((a-rec (get-rec a)) + (b-rec (get-rec b))) + ;; (print "a=" a " a-rec=" a-rec " host-last-used=" (host-last-used a-rec)) + ;; (print "b=" b " b-rec=" b-rec " host-last-used=" (host-last-used b-rec)) + (< (host-last-used a-rec) + (host-last-used b-rec)))))))) + (debug:print 0 *default-log-port* "INFO: hosts-sorted=" (get-hosts-sorted hosts)) + (if (null? hosts) + #f ;; no hosts to select from. All done and giving up now. + (let ((hosts-sorted (get-hosts-sorted hosts))) + (common:update-host-loads-table hosts) + (let loop ((hostname (car hosts-sorted)) + (tal (cdr hosts-sorted)) + (best-host #f)) + (let* ((rec (get-rec hostname)) + (reachable (host-reachable rec)) + (load (host-last-cpuload rec)) + (last-used (host-last-used rec)) + (delta (- curr-time last-used)) + (job-rate (if (> delta 0) + (/ 1 delta) + 999)) ;; jobs per second + (new-best + (cond + ((not reachable) + (debug:print 0 *default-log-port* "Skipping host " hostname " as it cannot be reached.") + best-host) + ((and (< load maxnload) ;; load is acceptable + (< job-rate maxjobrate)) ;; job rate is acceptable + (set! best-load load) + hostname) + (else best-host)))) + (debug:print 0 *default-log-port* "INFO: Trying host " hostname " with load " load ", last used " delta " seconds ago, with job-rate " job-rate " for running a test." ) + (if new-best + (begin ;; found a host, return it + (debug:print 0 *default-log-port* "INFO: Found host: " new-best " load: " load " last-used: " delta " seconds ago, with job-rate: " job-rate) + (host-last-used-set! rec curr-time) + new-best) + (if (null? tal) #f (loop (car tal)(cdr tal) best-host))))))))) + +(define (common:wait-for-cpuload maxload-in numcpus-in waitdelay #!key (count 1000) (msg #f)(remote-host #f)(force-maxload #f)) + (let* ((loadavg (common:get-cpu-load remote-host)) + (numcpus (if (<= 1 numcpus-in) ;; not possible to have zero. If we get 1, it's possible that we got the previous default, and we should check again + (common:get-num-cpus remote-host) + numcpus-in)) + (maxload (if force-maxload + maxload-in + (max maxload-in 0.5))) ;; so maxload must be greater than 0.5 for now BUG - FIXME? + (first (car loadavg)) + (next (cadr loadavg)) + (adjload (* maxload (max 1 numcpus))) ;; possible bug where numcpus (or could be maxload) is zero, crude fallback is to at least use 1 + (loadjmp (- first next)) + (adjwait (min (+ 300 (random 10)) (abs (* (+ (random 10)(/ (- 1000 count) 10) waitdelay) (- first adjload) )) ))) ;; add some randomness to the time to break any alignment where netbatch dumps many jobs to machines simultaneously + (debug:print-info 1 *default-log-port* "Checking cpuload on " (or remote-host "localhost") ", maxload: " maxload + ", load: " first ", adjload: " adjload ", loadjmp: " loadjmp) + (cond + ((and (> first adjload) + (> count 0)) + (debug:print-info 0 *default-log-port* "server start delayed " adjwait " seconds due to load " first " exceeding max of " adjload " on server " (or remote-host (get-host-name)) " (normalized load-limit: " maxload ") " (if msg msg "")) + (thread-sleep! adjwait) + (common:wait-for-cpuload maxload numcpus waitdelay count: (- count 1) msg: msg remote-host: remote-host)) + ((and (> loadjmp numcpus) + (> count 0)) + (debug:print-info 0 *default-log-port* "waiting " adjwait " seconds due to load jump " loadjmp " > numcpus " numcpus (if msg msg "")) + (thread-sleep! adjwait) + (common:wait-for-cpuload maxload numcpus waitdelay count: (- count 1) msg: msg remote-host: remote-host))))) + +(define (common:wait-for-homehost-load maxload msg) + (let* ((hh-dat (if (common:on-homehost?) ;; if we are on the homehost then pass in #f so the calls are local. + #f + (common:get-homehost))) + (hh (if hh-dat (car hh-dat) #f)) + (numcpus (common:get-num-cpus hh))) + (common:wait-for-normalized-load maxload msg hh))) + +(define (common:get-num-cpus remote-host) + (let* ((actual-host (or remote-host (get-host-name)))) + (or (common:get-cached-info actual-host "num-cpus" age: 86400) ;; hosts had better not be changing the number of cpus too often! + (let* ((proc (lambda () + (let loop ((numcpu 0) + (inl (read-line))) + (if (eof-object? inl) + (begin + (common:write-cached-info remote-host "num-cpus" numcpu) + numcpu) + (loop (if (string-match "^processor\\s+:\\s+\\d+$" inl) + (+ numcpu 1) + numcpu) + (read-line)))))) + (result (if remote-host + (with-input-from-pipe + (conc "ssh " remote-host " cat /proc/cpuinfo") + proc) + (with-input-from-file "/proc/cpuinfo" proc)))) + (common:write-cached-info actual-host "num-cpus" result) + result)))) + +;; wait for normalized cpu load to drop below maxload +;; +(define (common:wait-for-normalized-load maxload msg remote-host) + (let ((num-cpus (common:get-num-cpus remote-host))) + (common:wait-for-cpuload maxload num-cpus 15 msg: msg remote-host: remote-host))) + +;;====================================================================== +;; D E B U G G I N G S T U F F +;;====================================================================== + +;; (define *verbosity* 1) +;; (define *logging* #f) + +(define (common:set-last-run-version) + (rmt:set-var "MEGATEST_VERSION" (common:version-signature))) + +;; postive number if megatest version > db version +;; negative number if megatest version < db version +(define (common:version-db-delta) + (- megatest-version (common:get-last-run-version-number))) + +(define (common:version-changed?) + (not (equal? (common:get-last-run-version) + (common:version-signature)))) + +;; from metadat lookup MEGATEST_VERSION +;; +(define (common:get-last-run-version) ;; RADT => How does this work in send-receive function??; assume it is the value saved in some DB + (rmt:get-var "MEGATEST_VERSION")) + +(define (common:get-last-run-version-number) + (string->number + (substring (common:get-last-run-version) 0 6))) + +(define (common:api-changed?) + (not (equal? (substring (->string megatest-version) 0 4) + (substring (conc (common:get-last-run-version)) 0 4)))) + +;; '(print (string-intersperse (map cadr (hash-table-ref/default (read-config "megatest.config" \#f \#t) "disks" '"'"'("none" ""))) "\n"))' +(define (common:get-disks #!key (configf #f)) + (hash-table-ref/default + (or configf (configf:read-config "megatest.config" #f #t)) + "disks" '("none" ""))) + +;;====================================================================== +;; watchdog and exit procedures +;;====================================================================== + +;;====================================================================== +;; E X I T H A N D L I N G +;;====================================================================== + +(define (common:run-sync?) + (and (common:on-homehost?) + (args:get-arg "-server"))) + +;; (let ((ohh (common:on-homehost?)) +;; (srv (args:get-arg "-server"))) +;; (and ohh srv))) + ;; (debug:print-info 0 *default-log-port* "common:run-sync? ohh=" ohh ", srv=" srv) + +(define *watchdog* (make-thread + (lambda () + (handle-exceptions + exn + (begin + (print-call-chain) + (print " message: " ((condition-property-accessor 'exn 'message) exn))) + (common:watchdog))) + "Watchdog thread")) + +;; currently the primary job of the watchdog is to run the sync back to megatest.db from the db in /tmp +;; if we are on the homehost and we are a server (by definition we are on the homehost if we are a server) +;; +(define (common:readonly-watchdog dbstruct) + (thread-sleep! 0.05) ;; delay for startup + (debug:print-info 13 *default-log-port* "common:readonly-watchdog entered.") + ;; sync megatest.db to /tmp/.../megatst.db + (let* ((sync-cool-off-duration 3) + (golden-mtdb (dbr:dbstruct-mtdb dbstruct)) + (golden-mtpath (db:dbdat-get-path golden-mtdb)) + (tmp-mtdb (dbr:dbstruct-tmpdb dbstruct)) + (tmp-mtpath (db:dbdat-get-path tmp-mtdb))) + (debug:print-info 0 *default-log-port* "Read-only periodic sync thread started.") + (let loop ((last-sync-time 0)) + (debug:print-info 13 *default-log-port* "loop top tmp-mtpath="tmp-mtpath" golden-mtpath="golden-mtpath) + (let* ((duration-since-last-sync (- (current-seconds) last-sync-time))) + (debug:print-info 13 *default-log-port* "duration-since-last-sync="duration-since-last-sync) + (if (and (not *time-to-exit*) + (< duration-since-last-sync sync-cool-off-duration)) + (thread-sleep! (- sync-cool-off-duration duration-since-last-sync))) + (if (not *time-to-exit*) + (let ((golden-mtdb-mtime (file-modification-time golden-mtpath)) + (tmp-mtdb-mtime (file-modification-time tmp-mtpath))) + (if (> golden-mtdb-mtime tmp-mtdb-mtime) + (if (< golden-mtdb-mtime (- (current-seconds) 3)) ;; file has NOT been touched in past three seconds, this way multiple servers won't fight to sync back + (let ((res (db:multi-db-sync dbstruct 'old2new))) + (debug:print-info 13 *default-log-port* "rosync called, " res " records transferred.")))) + (loop (current-seconds))) + #t))) + (debug:print-info 0 *default-log-port* "Exiting readonly-watchdog timer, *time-to-exit* = " *time-to-exit*" pid="(current-process-id)" mtpath="golden-mtpath))) + + +;; TODO: for multiple areas, we will have multiple watchdogs; and multiple threads to manage +(define (common:watchdog) + (debug:print-info 13 *default-log-port* "common:watchdog entered.") + (if (launch:setup) + (if (common:on-homehost?) + (let ((dbstruct (db:setup #t))) + (debug:print-info 13 *default-log-port* "after db:setup with dbstruct=" dbstruct) + (cond + ((dbr:dbstruct-read-only dbstruct) + (debug:print-info 13 *default-log-port* "loading read-only watchdog") + (common:readonly-watchdog dbstruct)) + (else + (debug:print-info 13 *default-log-port* "loading writable-watchdog.") + (let* ((syncer (or (configf:lookup *configdat* "server" "sync-method") "brute-force-sync"))) + (cond + ((equal? syncer "brute-force-sync") + (server:writable-watchdog-bruteforce dbstruct)) + ((equal? syncer "delta-sync") + (server:writable-watchdog-deltasync dbstruct)) + (else + (debug:print-error 0 *default-log-port* "Unknown server/sync-method specified ("syncer") - valid values are brute-force-sync and delta-sync.") + (exit 1))) + ;;(debug:print 1 *default-log-port* "INFO: ["(common:human-time)"] Syncer started (method="syncer")") + ))) + (debug:print-info 13 *default-log-port* "watchdog done.")) + (debug:print-info 13 *default-log-port* "no need for watchdog on non-homehost")))) + + +(define (std-exit-procedure) + ;;(common:telemetry-log-close) + (on-exit (lambda () 0)) + ;;(debug:print-info 13 *default-log-port* "std-exit-procedure called; *time-to-exit*="*time-to-exit*) + (let ((no-hurry (if *time-to-exit* ;; hurry up + #f + (begin + (set! *time-to-exit* #t) + #t)))) + (debug:print-info 4 *default-log-port* "starting exit process, finalizing databases.") + (if (and no-hurry (debug:debug-mode 18)) + (rmt:print-db-stats)) + (let ((th1 (make-thread (lambda () ;; thread for cleaning up, give it five seconds + (if *dbstruct-db* (db:close-all *dbstruct-db*)) ;; one second allocated + (if *task-db* + (let ((db (cdr *task-db*))) + (if (sqlite3:database? db) + (begin + (sqlite3:interrupt! db) + (sqlite3:finalize! db #t) + ;; (vector-set! *task-db* 0 #f) + (set! *task-db* #f))))) + (http-client#close-all-connections!) + ;; (if (and *runremote* + ;; (remote-conndat *runremote*)) + ;; (begin + ;; (http-client#close-all-connections!))) ;; for http-client + (if (not (eq? *default-log-port* (current-error-port))) + (close-output-port *default-log-port*)) + (set! *default-log-port* (current-error-port))) "Cleanup db exit thread")) + (th2 (make-thread (lambda () + (debug:print 4 *default-log-port* "Attempting clean exit. Please be patient and wait a few seconds...") + (if no-hurry + (begin + (thread-sleep! 5)) ;; give the clean up few seconds to do it's stuff + (begin + (thread-sleep! 2))) + (debug:print 4 *default-log-port* " ... done") + ) + "clean exit"))) + (thread-start! th1) + (thread-start! th2) + (thread-join! th1) + ) + ) + + 0) + +(define (std-signal-handler signum) + ;; (signal-mask! signum) + (set! *time-to-exit* #t) + ;;(debug:print-info 13 *default-log-port* "got signal "signum) + (debug:print-error 0 *default-log-port* "Received signal " signum " aaa exiting promptly") + ;; (std-exit-procedure) ;; shouldn't need this since we are exiting and it will be called anyway + (exit)) + +(define (special-signal-handler signum) + ;; (signal-mask! signum) + (set! *time-to-exit* #t) + ;;(debug:print-info 13 *default-log-port* "got signal "signum) + (debug:print-error 0 *default-log-port* "Received signal " signum " sending email befor exiting!!") + ;;TODO send email to notify admin contact listed in the config that the lisner got killed + ;; (std-exit-procedure) ;; shouldn't need this since we are exiting and it will be called anyway + (exit)) + + +(set-signal-handler! signal/int std-signal-handler) ;; ^C +(set-signal-handler! signal/term std-signal-handler) + +;; (set-signal-handler! signal/stop std-signal-handler) ;; ^Z NO, do NOT handle ^Z! + +;; Force a megatest cleanup-db if version is changed and skip-version-check not specified +;; Do NOT check if not on homehost! +;; +(define (common:exit-on-version-changed) + (if (common:on-homehost?) + (if (common:api-changed?) + (let* ((mtconf (conc (get-environment-variable "MT_RUN_AREA_HOME") "/megatest.config")) + (dbfile (conc (get-environment-variable "MT_RUN_AREA_HOME") "/megatest.db")) + (read-only (not (file-write-access? dbfile))) + (dbstruct (db:setup #t))) + (debug:print 0 *default-log-port* + "WARNING: Version mismatch!\n" + " expected: " (common:version-signature) "\n" + " got: " (common:get-last-run-version)) + (cond + ((get-environment-variable "MT_SKIP_DB_MIGRATE") #t) + ((and (common:file-exists? mtconf) (common:file-exists? dbfile) (not read-only) + (eq? (current-user-id)(file-owner mtconf))) ;; safe to run -cleanup-db + (debug:print 0 *default-log-port* " I see you are the owner of megatest.config, attempting to cleanup and reset to new version") + (handle-exceptions + exn + (begin + (debug:print 0 *default-log-port* "Failed to switch versions.") + (debug:print 0 *default-log-port* " message: " ((condition-property-accessor 'exn 'message) exn)) + (print-call-chain (current-error-port)) + (exit 1)) + (common:cleanup-db dbstruct))) + ((not (common:file-exists? mtconf)) + (debug:print 0 *default-log-port* " megatest.config does not exist in this area. Cannot proceed with megatest version migration.") + (exit 1)) + ((not (common:file-exists? dbfile)) + (debug:print 0 *default-log-port* " megatest.db does not exist in this area. Cannot proceed with megatest version migration.") + (exit 1)) + ((not (eq? (current-user-id)(file-owner mtconf))) + (debug:print 0 *default-log-port* " You do not own megatest.db in this area. Cannot proceed with megatest version migration.") + (exit 1)) + (read-only + (debug:print 0 *default-log-port* " You have read-only access to this area. Cannot proceed with megatest version migration.") + (exit 1)) + (else + (debug:print 0 *default-log-port* " to switch versions you can run: \"megatest -cleanup-db\"") + (exit 1))))))) +;; (begin +;; (debug:print 0 *default-log-port* "ERROR: cannot migrate version unless on homehost. Exiting.") +;; (exit 1)))) +;; Move me elsewhere ... +;; RADT => Why do we meed the version check here, this is called only if version misma +;; +(define (common:cleanup-db dbstruct #!key (full #f)) + (apply db:multi-db-sync + dbstruct + 'schema + ;; 'new2old + 'killservers + 'adj-target + ;; 'old2new + 'new2old + ;; (if full + '(dejunk) + ;; '()) + ) + (if (common:api-changed?) + (common:set-last-run-version))) ) Index: rmtmod.scm ================================================================== --- rmtmod.scm +++ rmtmod.scm @@ -52,29 +52,10 @@ ;; db stats (define *db-stats* (make-hash-table)) ;; hash of vectors < count duration-total > (define *db-stats-mutex* (make-mutex)) - -(defstruct remote - (hh-dat (common:get-homehost)) ;; homehost record ( addr . hhflag ) - (server-url (if *toppath* (server:check-if-running *toppath*))) ;; (server:check-if-running *toppath*) #f)) - (last-server-check 0) ;; last time we checked to see if the server was alive - (conndat #f) - (transport *transport-type*) - (server-timeout (server:expiration-timeout)) - (force-server #f) - (ro-mode #f) - (ro-mode-checked #f)) ;; flag that indicates we have checked for ro-mode - -;; launching and hosts -(defstruct host - (reachable #f) - (last-update 0) - (last-used 0) - (last-cpuload 1)) - ;;====================================================================== ;; S U P P O R T F U N C T I O N S ;;====================================================================== Index: servermod.scm ================================================================== --- servermod.scm +++ servermod.scm @@ -18,17 +18,20 @@ ;;====================================================================== (declare (unit servermod)) (declare (uses commonmod)) +(declare (uses dbmod)) (module servermod * (import scheme chicken data-structures extras) (import (prefix sqlite3 sqlite3:) posix typed-records srfi-18 srfi-69 format ports srfi-1 matchable) (import commonmod) +(import dbmod) + ;; (use (prefix ulex ulex:)) (include "common_records.scm") (define (server:make-server-url hostport) @@ -36,10 +39,328 @@ #f (conc "http://" (car hostport) ":" (cadr hostport)))) (define *server-loop-heart-beat* (current-seconds)) +;; options: +;; +;; 'killservers - kills all servers +;; 'dejunk - removes junk records +;; 'adj-testids - move test-ids into correct ranges +;; 'old2new - sync megatest.db to /tmp/.../megatest.db and /tmp/.../megatest_ref.db +;; 'new2old - sync /tmp/.../megatest.db to megatest.db and /tmp/.../megatest_ref.db (and update data_synced) +;; 'closeall - close all opened dbs +;; 'schema - attempt to apply schema changes +;; run-ids: '(1 2 3 ...) or #f (for all) +;; +(define (db:multi-db-sync dbstruct . options) + ;; (if (not (launch:setup)) + ;; (debug:print 0 *default-log-port* "ERROR: not able to setup up for megatest.") + (let* ((mtdb (dbr:dbstruct-mtdb dbstruct)) + (tmpdb (db:get-db dbstruct)) + (refndb (dbr:dbstruct-refndb dbstruct)) + (allow-cleanup #t) ;; (if run-ids #f #t)) + (servers (server:get-list *toppath*)) ;; (tasks:get-all-servers (db:delay-if-busy tdbdat))) + (data-synced 0)) ;; count of changed records (I hope) + + (for-each + (lambda (option) + + (case option + ;; kill servers + ((killservers) + (for-each + (lambda (server) + (match-let (((mod-time host port start-time pid) server)) + (if (and host pid) + (tasks:kill-server host pid)))) + servers) + + ;; /tmp/bjbarcla/megatest_localdb/fullrun/.nfs.pdx.disks.icf_env_disk001.bjbarcla.gwa.issues.mtdev.matt-bisect.megatest.ext-tests.runs.all.v1.65.1.6524.dev.bb-24-justrollup-f8.rollup.fullrun/megatest.db.lock + (delete-file* (common:get-sync-lock-filepath)) + ) + + ;; clear out junk records + ;; + ((dejunk) + (db:delay-if-busy mtdb) ;; ok to delay on mtdb + (when (file-write-access? (db:dbdat-get-path mtdb)) (db:clean-up mtdb)) + (db:clean-up tmpdb) + (db:clean-up refndb)) + + ;; sync runs, test_meta etc. + ;; + ((old2new) + (set! data-synced + (+ (db:sync-tables (db:sync-all-tables-list dbstruct) #f mtdb tmpdb refndb) + data-synced))) + + ;; now ensure all newdb data are synced to megatest.db + ;; do not use the run-ids list passed in to the function + ;; + ((new2old) + (set! data-synced + (+ (db:sync-tables (db:sync-all-tables-list dbstruct) #f tmpdb refndb mtdb) + data-synced))) + + ((adj-target) + (db:adj-target (db:dbdat-get-db mtdb)) + (db:adj-target (db:dbdat-get-db tmpdb)) + (db:adj-target (db:dbdat-get-db refndb))) + + ((schema) + (db:patch-schema-maindb (db:dbdat-get-db mtdb)) + (db:patch-schema-maindb (db:dbdat-get-db tmpdb)) + (db:patch-schema-maindb (db:dbdat-get-db refndb)) + (db:patch-schema-rundb (db:dbdat-get-db mtdb)) + (db:patch-schema-rundb (db:dbdat-get-db tmpdb)) + (db:patch-schema-rundb (db:dbdat-get-db refndb)))) + + (stack-push! (dbr:dbstruct-dbstack dbstruct) tmpdb)) + options) + data-synced)) + +(define (db:tmp->megatest.db-sync dbstruct last-update) + (let* ((mtdb (dbr:dbstruct-mtdb dbstruct)) + (tmpdb (db:get-db dbstruct)) + (refndb (dbr:dbstruct-refndb dbstruct)) + (res (db:sync-tables (db:sync-all-tables-list dbstruct) last-update tmpdb refndb mtdb))) + (stack-push! (dbr:dbstruct-dbstack dbstruct) tmpdb) + res)) + +;;;; run-ids +;; if #f use *db-local-sync* : or 'local-sync-flags +;; if #t use timestamps : or 'timestamps +;; +;; NB// no-sync-db is the db handle, not a flag! +;; +(define (db:sync-to-megatest.db dbstruct #!key (no-sync-db #f)) + (let* ((start-time (current-seconds)) + (last-full-update (if no-sync-db + (db:no-sync-get/default no-sync-db "LAST_FULL_UPDATE" 0) + 0)) + (full-sync-needed (> (- start-time last-full-update) 3600)) ;; every hour do a full sync + (last-update (if full-sync-needed + 0 + (if no-sync-db + (db:no-sync-get/default no-sync-db "LAST_UPDATE" 0) + 0))) ;; (or (db:get-var dbstruct "LAST_UPDATE") 0)) + (sync-needed (> (- start-time last-update) 6)) + (res (if (or sync-needed ;; don't sync if a sync already occurred in the past 6 seconds + full-sync-needed) + (begin + (if no-sync-db + (begin + (if full-sync-needed (db:no-sync-set no-sync-db "LAST_FULL_UPDATE" start-time)) + (db:no-sync-set no-sync-db "LAST_UPDATE" start-time))) + (db:tmp->megatest.db-sync dbstruct last-update)) + 0)) + (sync-time (- (current-seconds) start-time))) + (debug:print-info 3 *default-log-port* "Sync of newdb to olddb completed in " sync-time " seconds pid="(current-process-id)) + (if (common:low-noise-print 30 "sync new to old") + (if sync-needed + (debug:print-info 0 *default-log-port* "Sync of " res " records from newdb to olddb completed in " sync-time " seconds pid="(current-process-id)) + (debug:print-info 0 *default-log-port* "No sync needed, last updated " (- start-time last-update) " seconds ago"))) + res)) + +;;====================================================================== +;; server stuff that operates on the server log files +;;====================================================================== + +(define (server:get-num-servers #!key (numservers 2)) + (let ((ns (string->number + (or (configf:lookup *configdat* "server" "numservers") "notanumber")))) + (or ns numservers))) + +;; get a list of servers with all relevant data +;; ( mod-time host port start-time pid ) +;; +(define (server:get-list areapath #!key (limit #f)) + (let ((fname-rx (regexp "^(|.*/)server-(\\d+)-(\\S+).log$")) + (day-seconds (* 24 60 60))) + ;; if the directory exists continue to get the list + ;; otherwise attempt to create the logs dir and then + ;; continue + (if (if (directory-exists? (conc areapath "/logs")) + '() + (if (file-write-access? areapath) + (begin + (condition-case + (create-directory (conc areapath "/logs") #t) + (exn (i/o file)(debug:print 0 *default-log-port* "ERROR: Cannot create directory at " (conc areapath "/logs"))) + (exn ()(debug:print 0 *default-log-port* "ERROR: Unknown error attemtping to get server list."))) + (directory-exists? (conc areapath "/logs"))) + '())) + (let* ((server-logs (glob (conc areapath "/logs/server-*.log"))) + (num-serv-logs (length server-logs))) + (if (null? server-logs) + '() + (let loop ((hed (car server-logs)) + (tal (cdr server-logs)) + (res '())) + (let* ((mod-time (handle-exceptions + exn + (current-seconds) ;; 0 + (file-modification-time hed))) ;; default to *very* old so log gets ignored if deleted + (down-time (- (current-seconds) mod-time)) + (serv-dat (if (or (< num-serv-logs 10) + (< down-time 900)) ;; day-seconds)) + (server:logf-get-start-info hed) + '())) ;; don't waste time processing server files not touched in the 15 minutes if there are more than ten servers to look at + (serv-rec (cons mod-time serv-dat)) + (fmatch (string-match fname-rx hed)) + (pid (if fmatch (string->number (list-ref fmatch 2)) #f)) + (new-res (if (null? serv-dat) + res + (cons (append serv-rec (list pid)) res)))) + (if (null? tal) + (if (and limit + (> (length new-res) limit)) + new-res ;; (take new-res limit) <= need intelligent sorting before this will work + new-res) + (loop (car tal)(cdr tal) new-res))))))))) +;; given a path to a server log return: host port startseconds +;; +(define (server:logf-get-start-info logf) + (let ((rx (regexp "^SERVER STARTED: (\\S+):(\\d+) AT ([\\d\\.]+)"))) ;; SERVER STARTED: host:port AT timesecs + (handle-exceptions + exn + (list #f #f #f) ;; no idea what went wrong, call it a bad server + (with-input-from-file + logf + (lambda () + (let loop ((inl (read-line)) + (lnum 0)) + (if (not (eof-object? inl)) + (let ((mlst (string-match rx inl))) + (if (not mlst) + (if (< lnum 500) ;; give up if more than 500 lines of server log read + (loop (read-line)(+ lnum 1)) + (list #f #f #f)) + (let ((dat (cdr mlst))) + (list (car dat) ;; host + (string->number (cadr dat)) ;; port + (string->number (caddr dat)))))) + (list #f #f #f)))))))) + +(define (server:get-num-alive srvlst) + (let ((num-alive 0)) + (for-each + (lambda (server) + (match-let (((mod-time host port start-time pid) + server)) + (let* ((uptime (- (current-seconds) mod-time)) + (runtime (if start-time + (- mod-time start-time) + 0))) + (if (< uptime 5)(set! num-alive (+ num-alive 1)))))) + srvlst) + num-alive)) + +;; given a list of servers get a list of valid servers, i.e. at least +;; 10 seconds old, has started and is less than 1 hour old and is +;; active (i.e. mod-time < 10 seconds +;; +;; mod-time host port start-time pid +;; +;; sort by start-time descending. I.e. get the oldest first. Young servers will thus drop off +;; and servers should stick around for about two hours or so. +;; +(define (server:get-best srvlst) + (let* ((nums (server:get-num-servers)) + (now (current-seconds)) + (slst (sort + (filter (lambda (rec) + (if (and (list? rec) + (> (length rec) 2)) + (let ((start-time (list-ref rec 3)) + (mod-time (list-ref rec 0))) + ;; (print "start-time: " start-time " mod-time: " mod-time) + (and start-time mod-time + (> (- now start-time) 0) ;; been running at least 0 seconds + (< (- now mod-time) 16) ;; still alive - file touched in last 16 seconds + (< (- now start-time) + (+ (- (string->number (or (configf:lookup *configdat* "server" "runtime") "3600")) + 180) + (random 360))) ;; under one hour running time +/- 180 + )) + #f)) + srvlst) + (lambda (a b) + (< (list-ref a 3) + (list-ref b 3)))))) + (if (> (length slst) nums) + (take slst nums) + slst))) + +(define (server:get-first-best areapath) + (let ((srvrs (server:get-best (server:get-list areapath)))) + (if (and srvrs + (not (null? srvrs))) + (car srvrs) + #f))) + +(define (server:get-rand-best areapath) + (let ((srvrs (server:get-best (server:get-list areapath)))) + (if (and (list? srvrs) + (not (null? srvrs))) + (let* ((len (length srvrs)) + (idx (random len))) + (list-ref srvrs idx)) + #f))) + + +(define (server:record->url servr) + (match-let (((mod-time host port start-time pid) + servr)) + (if (and host port) + (conc host ":" port) + #f))) + + +;; timeout is hms string: 1h 5m 3s, default is 1 minute +;; +(define (server:expiration-timeout) + (let ((tmo (configf:lookup *configdat* "server" "timeout"))) + (if (and (string? tmo) + (common:hms-string->seconds tmo)) ;; BUG: hms-string->seconds is broken, if given "10" returns 0. Also, it doesn't belong in this logic unless the string->number is changed below + (* 3600 (string->number tmo)) + 60))) + +;; ping the given server +;; +(define (server:check-server server-record) + (let* ((server-url (server:record->url server-record)) + (res (case *transport-type* + ((http)(server:ping server-url)) + ;; ((nmsg)(nmsg-transport:ping (tasks:hostinfo-get-interface server) + ))) + (if res + server-url + #f))) + +;; no longer care if multiple servers are started by accident. older servers will drop off in time. +;; +(define (server:check-if-running areapath) ;; #!key (numservers "2")) + (let* ((ns (server:get-num-servers)) + (servers (server:get-best (server:get-list areapath)))) + ;; (print "servers: " servers " ns: " ns) + (if (or (and servers + (null? servers)) + (not servers) + (and (list? servers) + (< (length servers) (random ns)))) ;; somewhere between 0 and numservers + #f + (let loop ((hed (car servers)) + (tal (cdr servers))) + (let ((res (server:check-server hed))) + (if res + res + (if (null? tal) + #f + (loop (car tal)(cdr tal))))))))) + ;;====================================================================== ;; P K T S S T U F F ;;====================================================================== ;; ??? @@ -197,43 +518,10 @@ (loop (server:check-if-running areapath) (+ try-num 1))))))) (define server:try-running server:run) ;; there is no more per-run servers ;; REMOVE ME. BUG. -;; no longer care if multiple servers are started by accident. older servers will drop off in time. -;; -(define (server:check-if-running areapath) ;; #!key (numservers "2")) - (let* ((ns (server:get-num-servers)) - (servers (server:get-best (server:get-list areapath)))) - ;; (print "servers: " servers " ns: " ns) - (if (or (and servers - (null? servers)) - (not servers) - (and (list? servers) - (< (length servers) (random ns)))) ;; somewhere between 0 and numservers - #f - (let loop ((hed (car servers)) - (tal (cdr servers))) - (let ((res (server:check-server hed))) - (if res - res - (if (null? tal) - #f - (loop (car tal)(cdr tal))))))))) - -;; ping the given server -;; -(define (server:check-server server-record) - (let* ((server-url (server:record->url server-record)) - (res (case *transport-type* - ((http)(server:ping server-url)) - ;; ((nmsg)(nmsg-transport:ping (tasks:hostinfo-get-interface server) - ))) - (if res - server-url - #f))) - (define (server:kill servr) (match-let (((mod-time hostname port start-time pid) servr)) (tasks:kill-server hostname pid))) @@ -303,32 +591,10 @@ (set! *db-last-access* (current-seconds)) ;; might not be needed. (if (equal? *toppath* toppath) #t #f))) -;; timeout is hms string: 1h 5m 3s, default is 1 minute -;; -(define (server:expiration-timeout) - (let ((tmo (configf:lookup *configdat* "server" "timeout"))) - (if (and (string? tmo) - (common:hms-string->seconds tmo)) ;; BUG: hms-string->seconds is broken, if given "10" returns 0. Also, it doesn't belong in this logic unless the string->number is changed below - (* 3600 (string->number tmo)) - 60))) - -(define (server:get-best-guess-address hostname) - (let ((res #f)) - (for-each - (lambda (adr) - (if (not (eq? (u8vector-ref adr 0) 127)) - (set! res adr))) - ;; NOTE: This can fail when there is no mention of the host in /etc/hosts. FIXME - (vector->list (hostinfo-addresses (hostname->hostinfo hostname)))) - (string-intersperse - (map number->string - (u8vector->list - (if res res (hostname->ip hostname)))) "."))) - ;; (define server:sync-lock-token "SERVER_SYNC_LOCK") ;; (define (server:release-sync-lock) ;; (db:no-sync-del! *no-sync-db* server:sync-lock-token)) ;; (define (server:have-sync-lock?) ;; (let* ((have-lock-pair (db:no-sync-get-lock *no-sync-db* server:sync-lock-token)) Index: tasksmod.scm ================================================================== --- tasksmod.scm +++ tasksmod.scm @@ -286,165 +286,10 @@ (set! res count)) mdb "SELECT count(id) FROM monitors WHERE last_update < (strftime('%s','now') - 300) AND username=?;" (car (user-information (current-user-id)))) res)) - -;;====================================================================== -;; server stuff that operates on the server log files -;;====================================================================== - -(define (server:get-num-servers #!key (numservers 2)) - (let ((ns (string->number - (or (configf:lookup *configdat* "server" "numservers") "notanumber")))) - (or ns numservers))) - -;; get a list of servers with all relevant data -;; ( mod-time host port start-time pid ) -;; -(define (server:get-list areapath #!key (limit #f)) - (let ((fname-rx (regexp "^(|.*/)server-(\\d+)-(\\S+).log$")) - (day-seconds (* 24 60 60))) - ;; if the directory exists continue to get the list - ;; otherwise attempt to create the logs dir and then - ;; continue - (if (if (directory-exists? (conc areapath "/logs")) - '() - (if (file-write-access? areapath) - (begin - (condition-case - (create-directory (conc areapath "/logs") #t) - (exn (i/o file)(debug:print 0 *default-log-port* "ERROR: Cannot create directory at " (conc areapath "/logs"))) - (exn ()(debug:print 0 *default-log-port* "ERROR: Unknown error attemtping to get server list."))) - (directory-exists? (conc areapath "/logs"))) - '())) - (let* ((server-logs (glob (conc areapath "/logs/server-*.log"))) - (num-serv-logs (length server-logs))) - (if (null? server-logs) - '() - (let loop ((hed (car server-logs)) - (tal (cdr server-logs)) - (res '())) - (let* ((mod-time (handle-exceptions - exn - (current-seconds) ;; 0 - (file-modification-time hed))) ;; default to *very* old so log gets ignored if deleted - (down-time (- (current-seconds) mod-time)) - (serv-dat (if (or (< num-serv-logs 10) - (< down-time 900)) ;; day-seconds)) - (server:logf-get-start-info hed) - '())) ;; don't waste time processing server files not touched in the 15 minutes if there are more than ten servers to look at - (serv-rec (cons mod-time serv-dat)) - (fmatch (string-match fname-rx hed)) - (pid (if fmatch (string->number (list-ref fmatch 2)) #f)) - (new-res (if (null? serv-dat) - res - (cons (append serv-rec (list pid)) res)))) - (if (null? tal) - (if (and limit - (> (length new-res) limit)) - new-res ;; (take new-res limit) <= need intelligent sorting before this will work - new-res) - (loop (car tal)(cdr tal) new-res))))))))) -;; given a path to a server log return: host port startseconds -;; -(define (server:logf-get-start-info logf) - (let ((rx (regexp "^SERVER STARTED: (\\S+):(\\d+) AT ([\\d\\.]+)"))) ;; SERVER STARTED: host:port AT timesecs - (handle-exceptions - exn - (list #f #f #f) ;; no idea what went wrong, call it a bad server - (with-input-from-file - logf - (lambda () - (let loop ((inl (read-line)) - (lnum 0)) - (if (not (eof-object? inl)) - (let ((mlst (string-match rx inl))) - (if (not mlst) - (if (< lnum 500) ;; give up if more than 500 lines of server log read - (loop (read-line)(+ lnum 1)) - (list #f #f #f)) - (let ((dat (cdr mlst))) - (list (car dat) ;; host - (string->number (cadr dat)) ;; port - (string->number (caddr dat)))))) - (list #f #f #f)))))))) - -(define (server:get-num-alive srvlst) - (let ((num-alive 0)) - (for-each - (lambda (server) - (match-let (((mod-time host port start-time pid) - server)) - (let* ((uptime (- (current-seconds) mod-time)) - (runtime (if start-time - (- mod-time start-time) - 0))) - (if (< uptime 5)(set! num-alive (+ num-alive 1)))))) - srvlst) - num-alive)) - -;; given a list of servers get a list of valid servers, i.e. at least -;; 10 seconds old, has started and is less than 1 hour old and is -;; active (i.e. mod-time < 10 seconds -;; -;; mod-time host port start-time pid -;; -;; sort by start-time descending. I.e. get the oldest first. Young servers will thus drop off -;; and servers should stick around for about two hours or so. -;; -(define (server:get-best srvlst) - (let* ((nums (server:get-num-servers)) - (now (current-seconds)) - (slst (sort - (filter (lambda (rec) - (if (and (list? rec) - (> (length rec) 2)) - (let ((start-time (list-ref rec 3)) - (mod-time (list-ref rec 0))) - ;; (print "start-time: " start-time " mod-time: " mod-time) - (and start-time mod-time - (> (- now start-time) 0) ;; been running at least 0 seconds - (< (- now mod-time) 16) ;; still alive - file touched in last 16 seconds - (< (- now start-time) - (+ (- (string->number (or (configf:lookup *configdat* "server" "runtime") "3600")) - 180) - (random 360))) ;; under one hour running time +/- 180 - )) - #f)) - srvlst) - (lambda (a b) - (< (list-ref a 3) - (list-ref b 3)))))) - (if (> (length slst) nums) - (take slst nums) - slst))) - -(define (server:get-first-best areapath) - (let ((srvrs (server:get-best (server:get-list areapath)))) - (if (and srvrs - (not (null? srvrs))) - (car srvrs) - #f))) - -(define (server:get-rand-best areapath) - (let ((srvrs (server:get-best (server:get-list areapath)))) - (if (and (list? srvrs) - (not (null? srvrs))) - (let* ((len (length srvrs)) - (idx (random len))) - (list-ref srvrs idx)) - #f))) - - -(define (server:record->url servr) - (match-let (((mod-time host port start-time pid) - servr)) - (if (and host port) - (conc host ":" port) - #f))) - ;; #;(define (tasks:start-monitor db mdb) (if (> (tasks:get-num-alive-monitors mdb) 2) ;; have two running, no need for more (debug:print-info 1 *default-log-port* "Not starting monitor, already have more than two running") Index: testsmod.scm ================================================================== --- testsmod.scm +++ testsmod.scm @@ -416,82 +416,7 @@ (if (null? tal) (begin (debug:print-error 0 *default-log-port* "No readable testconfig found for " test-name) #f) (loop (car tal)(cdr tal)))))))))) - -;;====================================================================== -;; T E S T L A U N C H I N G P E R I T E M W I T H H O S T T Y P E S -;;====================================================================== -;; -;; [hosts] -;; arm cubie01 cubie02 -;; x86_64 zeus xena myth01 -;; allhosts #{g hosts arm} #{g hosts x86_64} -;; -;; [host-types] -;; general #MTLOWESTLOAD #{g hosts allhosts} -;; arm #MTLOWESTLOAD #{g hosts arm} -;; nbgeneral nbjob run JOBCOMMAND -log $MT_LINKTREE/$MT_TARGET/$MT_RUNNAME.$MT_TESTNAME-$MT_ITEM_PATH.lgo -;; -;; [host-rules] -;; # maxnload => max normalized load -;; # maxnjobs => max jobs per cpu -;; # maxjobrate => max jobs per second -;; general maxnload=1.1; maxnjobs=1.2; maxjobrate=0.1 -;; -;; [launchers] -;; envsetup general -;; xor/%/n 4C16G -;; % nbgeneral -;; -;; [jobtools] -;; # if defined and not "no" flexi-launcher will bypass "launcher" unless no match. -;; flexi-launcher yes -;; launcher nbfake -;; -(define (common:get-launcher configdat testname itempath) - (let ((fallback-launcher (configf:lookup configdat "jobtools" "launcher"))) - (if (and (configf:lookup configdat "jobtools" "flexi-launcher") ;; overrides launcher - (not (equal? (configf:lookup configdat "jobtools" "flexi-launcher") "no"))) - (let* ((launchers (hash-table-ref/default configdat "launchers" '()))) - (if (null? launchers) - fallback-launcher - (let loop ((hed (car launchers)) - (tal (cdr launchers))) - (let ((patt (car hed)) - (host-type (cadr hed))) - (if (tests:match patt testname itempath) - (begin - (debug:print-info 2 *default-log-port* "Have flexi-launcher match for " testname "/" itempath " = " host-type) - (let ((launcher (configf:lookup configdat "host-types" host-type))) - (if launcher - (let* ((launcher-parts (string-split launcher)) - (launcher-exe (car launcher-parts))) - (if (equal? launcher-exe "#MTLOWESTLOAD") ;; this is our special case, we will find the lowest load and craft a nbfake commandline - (let host-loop ((targ-host (common:get-least-loaded-host (cdr launcher-parts) host-type configdat)) - (count 100)) - (if targ-host - (conc "remrun " targ-host) - (if (> count 0) - (begin - (debug:print 0 *default-log-port* "INFO: Waiting for a host for host-type " host-type) - (thread-sleep! (- 101 count)) - (host-loop (common:get-least-loaded-host (cdr launcher-parts) host-type configdat) - (- count 1))) - (begin - (debug:print 0 *default-log-port* "FATAL: Failed to find a host from #MTLOWESTLOAD for host-type " host-type) - (exit))))) - launcher)) - (begin - (debug:print-info 0 *default-log-port* "WARNING: no launcher found for host-type " host-type) - (if (null? tal) - fallback-launcher - (loop (car tal)(cdr tal))))))) - ;; no match, try again - (if (null? tal) - fallback-launcher - (loop (car tal)(cdr tal)))))))) - fallback-launcher))) - )