[tei-council] http://www.tei-c.org/oxgarage java.lang.OutOfMemoryError

[James Cummings]

On 17/02/14 17:42, Ian Rifkin wrote:
> Hi,
>
> I was able to SSH to the server as myself, but I couldn't switch
> to the TEI user. It reported that "cannot set user id: resource
> temporarily unavailable." Digging some more it didn't look like
> TEI was using a lot of processes on the server, but it was using
> a lot of open files. Perhaps there's something in the TEI code
> that doesn't close files properly after opening them, or perhaps
> there is just a lot of stuff happening on the site right now.

We'll try to investigate this.

> Either way, the server wasn't running out of power so I increased
> the soft limit from 1024 files to a maximum of 4096 files.
> Because of the state that TEI got in I had to kill TEI Tomcat and
> start it back up. I'm now able to login as the tei user, so I
> think things are back to normal.

I can now ssh in as well.


> Do you have a sense for how many open files TEI would need to
> have on the server? Do you think it's a bug in your code or that
> it just needs a higher limit and we should be all set?

No idea.
>
> Also, FYI, looking through today's email threads I see that there
> was some concern about being able to reach me -- and even John
> emailed me that  you were having trouble. Am I correct that the
> first email about this was at 8:30am EST today or is there an
> earlier email that I missed? TEI folks may not know the agreement
> with me: Basically I have a set number of hours per month that I
> will work for all sites on this server, which I typically will do
> during off hours (weekends, evenings). I agreed that I will
> respond "best effort" to outages, but there is no set expectation
> of response time. This was a particularly busy weekend (baby
> shower for my wife!) so I was not on email as much, but I think 4
> hour response time (and I think it's fixed) is reasonable. If you
> have different expectations there should  be a conversation.

erm, I emailed you on the Weds 12th Feb about this and got no 
response.

I understand and appreciate the best effort support arrangement 
-- I do a fair bit of that myself. I just dislike the idea of any 
system relying on one person. Us having a way some way to 
escalate to doing some of these things in your absence (but not 
doing so if you are around) seems like it might be beneficial.

Thanks for your help and for fixing the problem quickly.  I think 
we might need to revert to restarting tomcat regularly.

-James
-- 
Dr James Cummings, James.Cummings at it.ox.ac.uk
Academic IT Services, University of Oxford