Just set up god on a new server, running nginx with a mongrel
cluster. I'm getting this error in the log:
FATAL: Unhandled exception in driver loop - (Errno::ENOENT): No such
file or directory - /proc/5138/stat
/usr/local/lib/ruby/gems/1.8/gems/god-0.7.6/bin/../lib/god/system/
slash_proc_poller.rb:57:in `read'
/usr/local/lib/ruby/gems/1.8/gems/god-0.7.6/bin/../lib/god/system/
slash_proc_poller.rb:57:in `stat'
/usr/local/lib/ruby/gems/1.8/gems/god-0.7.6/bin/../lib/god/system/
slash_proc_poller.rb:28:in `percent_cpu'
/usr/local/lib/ruby/gems/1.8/gems/god-0.7.6/bin/../lib/god/system/
process.rb:27:in `percent_cpu'
/usr/local/lib/ruby/gems/1.8/gems/god-0.7.6/bin/../lib/god/conditions/
cpu_usage.rb:65:in `test'
/usr/local/lib/ruby/gems/1.8/gems/god-0.7.6/bin/../lib/god/task.rb:
317:in `handle_poll'
/usr/local/lib/ruby/gems/1.8/gems/god-0.7.6/bin/../lib/god/driver.rb:
68:in `handle_event'
I'm guessing that one of the mongrel processes has stopped but not
completely cleaned up after itself. any ideas on nudging god to
restart the process instead of blow up?
> Are you running FreeBSD 7.0? We just got a patch in that fixed
> detection of the proc file system when the directory exists but the fs
> isn't mounted:
[2008-06-18 13:44:21] FATAL: Unhandled exception in driver loop - (Errno::ENOENT): No such file or directory - /proc/20082/stat /usr/local/lib/ruby/gems/1.8/gems/god-0.7.7/bin/../lib/god/system/ slash_proc_poller.rb:71:in `read'
Is this possibly a configuration issue on my side?
Doh. It _could_ be configuration, but I don't think so.
When this occurs, is the process running? Seems like it could be a race condition where the process has died, but god hasn't noticed, so it goes to check stats in /proc, but the info for the now missing pid isn't there.
> Doh. It _could_ be configuration, but I don't think so.
> When this occurs, is the process running? Seems like it could be a > race condition where the process has died, but god hasn't noticed, so > it goes to check stats in /proc, but the info for the now missing pid > isn't there.
> On Wed, Jun 18, 2008 at 7:16 AM, Matt Cowley <madcow...@gmail.com> > wrote:
>> Hmm..got the same issue again:
>> [2008-06-18 13:44:21] FATAL: Unhandled exception in driver loop - >> (Errno::ENOENT): No such file or directory - /proc/20082/stat >> /usr/local/lib/ruby/gems/1.8/gems/god-0.7.7/bin/../lib/god/system/ >> slash_proc_poller.rb:71:in `read'
>> Is this possibly a configuration issue on my side?
>> thanks
>> On Jun 17, 2008, at 6:30 PM, Kevin Clark wrote:
>>> Tom just pushed 0.7.7 to Rubyforge. Try it out, let us know if it >>> fixes your issue.
>> On Jun 18, 2008, at 1:40 PM, Kevin Clark wrote:
>> Doh. It _could_ be configuration, but I don't think so.
>> When this occurs, is the process running? Seems like it could be a >> race condition where the process has died, but god hasn't noticed, so >> it goes to check stats in /proc, but the info for the now missing pid >> isn't there.
>On Wed, Jun 18, 2008 at 12:31 PM, Matt Cowley <madcow...@gmail.com> wrote:
> That sounds likely. It ran for several hours before this exception > occurred, so my guess is that it didn't get thrown until a process died.
Ok, I'll try to take a look today. In the meantime, if you want to modify your install so SlashProcPoller.usable? always returns false, it'll go back to using ps. This should probably be configurable anyway.
On Wed, Jun 18, 2008 at 12:55 PM, Kevin Clark <kevin.cl...@gmail.com> wrote: >>> On Jun 18, 2008, at 1:40 PM, Kevin Clark wrote:
>>> Doh. It _could_ be configuration, but I don't think so.
>>> When this occurs, is the process running? Seems like it could be a >>> race condition where the process has died, but god hasn't noticed, so >>> it goes to check stats in /proc, but the info for the now missing pid >>> isn't there.
>>On Wed, Jun 18, 2008 at 12:31 PM, Matt Cowley <madcow...@gmail.com> wrote:
>> That sounds likely. It ran for several hours before this exception >> occurred, so my guess is that it didn't get thrown until a process died.
> Ok, I'll try to take a look today. In the meantime, if you want to > modify your install so SlashProcPoller.usable? always returns false, > it'll go back to using ps. This should probably be configurable > anyway.
> I'm poking Tom about pushing an 0.7.9 release today to fix the issue.
> On Wed, Jun 18, 2008 at 12:55 PM, Kevin Clark > <kevin.cl...@gmail.com> wrote: >>>> On Jun 18, 2008, at 1:40 PM, Kevin Clark wrote:
>>>> Doh. It _could_ be configuration, but I don't think so.
>>>> When this occurs, is the process running? Seems like it could be a >>>> race condition where the process has died, but god hasn't >>>> noticed, so >>>> it goes to check stats in /proc, but the info for the now missing >>>> pid >>>> isn't there.
>>> On Wed, Jun 18, 2008 at 12:31 PM, Matt Cowley >>> <madcow...@gmail.com> wrote:
>>> That sounds likely. It ran for several hours before this exception >>> occurred, so my guess is that it didn't get thrown until a process >>> died.
>> Ok, I'll try to take a look today. In the meantime, if you want to >> modify your install so SlashProcPoller.usable? always returns false, >> it'll go back to using ps. This should probably be configurable >> anyway.
