[Buildbot-commits] [Buildbot] #1784: Buildmaster blocks

[Buildbot]

#1784: Buildmaster blocks
---------------------+--------------------
Reporter:  Schenker  |       Owner:
    Type:  defect    |      Status:  new
Priority:  major     |   Milestone:  0.8.4
 Version:  0.8.0     |  Resolution:
Keywords:            |
---------------------+--------------------

Comment (by Schenker):

 I ran strace on the master process. As I mentioned above - during the time
 when the master is frozen the web page cannot be opened and there's no
 activity marked in the twistd.log (I notice that there's no activity in
 the twistd.log because every 2 minutes the master checks the SVN for the
 latest revision and that is written in the log - when the master is frozen
 there's no such activity in the log). But in that time the strace shows
 output, including the SVN polling:


 {{{
 select(12, [3 5 6 8 9 11], [], [], {0, 687206}) = 0 (Timeout)
 write(4, "2011-02-03 14:07:00+0200 [-] SVN"..., 47) = 47
 pipe([12, 13])                          = 0
 pipe([14, 15])                          = 0
 pipe([16, 17])                          = 0
 clone(child_stack=0,
 flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD,
 child_tidptr=0x2ab3ffba5ff0) = 19642
 close(12)                               = 0
 fcntl(13, F_GETFL)                      = 0x1 (flags O_WRONLY)
 fcntl(13, F_SETFL, O_WRONLY|O_NONBLOCK) = 0
 fstat(13, {st_mode=S_IFIFO|0600, st_size=0, ...}) = 0
 close(15)                               = 0
 fcntl(14, F_GETFL)                      = 0 (flags O_RDONLY)
 fcntl(14, F_SETFL, O_RDONLY|O_NONBLOCK) = 0
 close(17)                               = 0
 fcntl(16, F_GETFL)                      = 0 (flags O_RDONLY)
 fcntl(16, F_SETFL, O_RDONLY|O_NONBLOCK) = 0
 wait4(19642, 0x7fffb3a25b64, WNOHANG, NULL) = 0
 select(17, [3 5 6 8 9 11 13 14 16], [], [], {0, 306986}) = 1 (in [14],
 left {0, 294000})
 read(14, "<?xml version=\"1.0\"?>\n<log>\n", 8192) = 28
 select(17, [3 5 6 8 9 11 13 14 16], [], [], {0, 293722}) = 1 (in [14],
 left {0, 232000})
 read(14, "<logentry\n   revision=\"37510\">\n<"..., 8192) = 1039
 select(17, [3 5 6 8 9 11 13 14 16], [], [], {0, 230957}) = 1 (in [14],
 left {0, 230957})
 read(14, "<logentry\n   revision=\"37501\">\n<"..., 8192) = 4016
 select(17, [3 5 6 8 9 11 13 14 16], [], [], {0, 230740}) = 1 (in [14],
 left {0, 230740})
 read(14, "<logentry\n   revision=\"37489\">\n<"..., 8192) = 1196
 select(17, [3 5 6 8 9 11 13 14 16], [], [], {0, 230546}) = 3 (in [13 14
 16], left {0, 228000})
 --- SIGCHLD (Child exited) @ 0 (0) ---
 write(10, "x", 1)                       = 1
 rt_sigreturn(0x2)                       = 3
 read(14, "", 8192)                      = 0
 close(14)                               = 0
 fcntl(13, F_GETFL)                      = 0x801 (flags
 O_WRONLY|O_NONBLOCK)
 fcntl(13, F_SETFL, O_WRONLY)            = 0
 close(13)                               = 0
 read(16, "", 8192)                      = 0
 close(16)                               = 0
 wait4(19642, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], WNOHANG, NULL) =
 19642
 write(4, "2011-02-03 14:07:00+0200 [-] svn"..., 85) = 85
 write(4, "2011-02-03 14:07:00+0200 [-] svn"..., 72) = 72
 write(4, "2011-02-03 14:07:00+0200 [-] SVN"..., 56) = 56
 write(4, "2011-02-03 14:07:00+0200 [-] _fi"..., 46) = 46
 select(12, [3 5 6 8 9 11], [], [], {0, 218745}) = 1 (in [3], left {0,
 218745})
 read(3, "x", 8192)                      = 1
 select(12, [3 5 6 8 9 11], [], [], {0, 218513}) = 0 (Timeout)
 select(12, [3 5 6 8 9 11], [], [], {0, 17}) = 0 (Timeout)
 select(12, [3 5 6 8 9 11], [], [], {0, 999948}) = 0 (Timeout)
 }}}

 So ... obviously the master is not entirely blocked ... but the result is
 the same - about 20 minutes after the master stops resonding the slave is
 detached and the build - interrupted with the following output:


 {{{
 [Failure instance: Traceback (failure with no frames): <class
 'twisted.internet.error.ConnectionLost'>: Connection to the other side was
 lost in a non-clean fashion.
 ]
 }}}

-- 
Ticket URL: <http://trac.buildbot.net/ticket/1784#comment:6>
Buildbot <http://buildbot.net/>
Buildbot: build/test automation