X-Git-Url: http://git.annexia.org/?p=whenjobs.git;a=blobdiff_plain;f=daemon%2Fdaemon.ml;h=42c32dee7bc10df44adbc2b1b7cb253d102d7cf3;hp=542c7a48b40ffb255708c766928ca430616aab82;hb=cffe87109d9c868eefd48f7d5255a4863a578e4d;hpb=21a3155f8296175180e348f10a6a9af88e1ac87d diff --git a/daemon/daemon.ml b/daemon/daemon.ml index 542c7a4..42c32de 100644 --- a/daemon/daemon.ml +++ b/daemon/daemon.ml @@ -32,8 +32,8 @@ let jobsdir = ref "" (* The state. *) let state = ref Whenstate.empty -(* Jobs that are running; map of PID -> (job, other data). Note that - * the job may no longer exist *OR* it may have been renamed, +(* Jobs that are running: a map of PID -> (job, tmpdir, serial). + * Note that the job may no longer exist *OR* it may have been renamed, * eg. if the jobs file was reloaded. *) let running = ref IntMap.empty @@ -292,18 +292,17 @@ and string_of_time_t t = tm.tm_hour tm.tm_min tm.tm_sec and run_job job = - let () = - (* Increment JOBSERIAL. *) - let serial = - match Whenstate.get_variable !state "JOBSERIAL" with - | T_int serial -> - let serial = succ_big_int serial in - state := Whenstate.set_variable !state "JOBSERIAL" (T_int serial); - serial - | _ -> assert false in - - Syslog.notice "running %s (JOBSERIAL=%s)" - job.job_name (string_of_big_int serial) in + (* Increment JOBSERIAL. *) + let serial = + match Whenstate.get_variable !state "JOBSERIAL" with + | T_int serial -> + let serial = succ_big_int serial in + state := Whenstate.set_variable !state "JOBSERIAL" (T_int serial); + serial + | _ -> assert false in + + Syslog.notice "running %s (JOBSERIAL=%s)" + job.job_name (string_of_big_int serial); (* Create a temporary directory. The current directory of the job * will be in this directory. The directory is removed when the @@ -324,7 +323,7 @@ and run_job job = putenv "JOBNAME" job.job_name; (* Create a temporary file containing the shell script fragment. *) - let script = dir // "script" in + let script = dir // "script.sh" in let chan = open_out script in fprintf chan "set -e\n"; (* So that jobs exit on error. *) output_string chan job.job_script.sh_script; @@ -333,6 +332,13 @@ and run_job job = let shell = try getenv "SHELL" with Not_found -> "/bin/sh" in + (* Set output to file. *) + let output = dir // "output.txt" in + let fd = openfile output [O_WRONLY; O_CREAT; O_TRUNC; O_NOCTTY] 0o600 in + dup2 fd stdout; + dup2 fd stderr; + close fd; + (* Execute the shell script. *) (try execvp shell [| shell; "-c"; script |]; with Unix_error (err, fn, _) -> @@ -344,7 +350,7 @@ and run_job job = (* Remember this PID, the job and the temporary directory, so we * can clean up when the child exits. *) - running := IntMap.add pid (job, dir) !running + running := IntMap.add pid (job, dir, serial) !running and tmpdir () = let chan = open_in "/dev/urandom" in @@ -362,13 +368,37 @@ and handle_sigchld _ = let pid, status = waitpid [WNOHANG] 0 in if pid > 0 then ( (* Look up the PID in the running jobs map. *) - let job, dir = IntMap.find pid !running in + let job, dir, serial = IntMap.find pid !running in running := IntMap.remove pid !running; - cleanup_job job dir + cleanup_job job dir serial status ) with Unix_error _ | Not_found -> () -and cleanup_job job dir = +and cleanup_job job dir serial status = + (* If there is a cleanup function, run it. *) + (match job.job_cleanup with + | None -> () + | Some cleanup -> + let code = + match status with + | WEXITED c -> c + | WSIGNALED s | WSTOPPED s -> 1 in + let result = { + res_job_name = job.job_name; + res_serial = serial; + res_code = code; + res_tmpdir = dir; + res_output = dir // "output.txt" + } in + try cleanup result + with + | Failure msg -> + Syslog.error "job %s cleanup function failed: %s" job.job_name msg + | exn -> + Syslog.error "job %s cleanup function exception: %s" + job.job_name (Printexc.to_string exn) + ); + (* This should be safe because the path cannot contain shell metachars. *) let cmd = sprintf "rm -rf '%s'" dir in ignore (Sys.command cmd)