Here is what I can tell you: 1) There are 4 managed repos 2) There are 8 remote repos. 3) At least one of the managed repos has over 200 GB. 4) There are single artifacts as big 100 MB. 5) There are test jars, source jars, and test-source jars which all cause the invalid path bug. 6) There is a single proxy connector for the 8 remote repos and this proxy is being used by our CI box as well as some developers so it is being hit on a continuous basis day and night. With fairly frequent and heavy load during the day. 7) Repository scans were setup for the 4 managed repos to run every 5 mins or so. Don't have the exact cron expressions anymore. This was adjusted to lengthen the time between scans and couple of repos were changed to only scan once a day. Once this was changed, the crash has not happened.
It definitely seems like a resource leak issue to me because the request would just hang and I have seen this behavior before. It should be an easy thing to profile. -----Original Message----- From: Brett Porter [mailto:[EMAIL PROTECTED] Sent: Sunday, February 24, 2008 2:26 PM To: archiva-users@maven.apache.org Subject: Re: Archiva crashes after a couple of days On 23/02/2008, Eric Miles <[EMAIL PROTECTED]> wrote: > Wow, you got us beat. Those are pretty large repos. Maybe the Archiva > team can comment on this. > > I'm interested to hear what they say. I run it on localhost with just around 500M which is all the stuff from central that I use on a daily basis. We've successfully run it on a copy of the central repository which I believe has a similar number of artifacts but is not as large. I have also run it on an 80G repo that was mostly very large files, so a smaller number of artifacts. We have got some reports of excessive memory use (which might cause this) over a large number of proxy requests and James has been investigating that recently. I think we can certainly resolve this problem with more investigation if it continues - but it really needs some more information on what is happening in that environment and narrowing down the possible causes. Cheers, Brett -- Brett Porter Blog: http://blogs.exist.com/bporter/
