On Tue, Aug 16, 2011 at 11:36:29AM -0400, Nico Kadel-Garcia wrote: > On Mon, Aug 15, 2011 at 11:08 PM, Seth Daniel > <subversion....@sethdaniel.org> wrote: > > Hi, > > > > I've done many subversion installs but never have I seen the following > > problem before. > > > > Vanilla Centos 5.4. Subversion 1.6.17 (compiled and packaged by me). > > Nothing fancy with the compilation. I use an 'amalgamated' sqlite > > 3.6.23 when configuring...otherwise no other options to configure. Then > > make; make install. The client being used is from the same 1.6.17 > > package. Apache with prefork MPM (centos 5.4 comes with apache 2.2.3). > > Dude!!! I'm building up an SRPM for RPMforge, backported from the > Fedora Core 16 test release of subversion-1.6.17. Would you care to > share notes,, or perhaps even test mine as scratch monkey > (http://www.catb.org/jargon/html/S/scratch-monkey.html) ? > > I also encourage you to jump to CentOS 5.6: while it took a while to > come out, there are numerous small improvements that pay off at tough > moments, like the profound stability updates to autofs. > > > > The problem: Occasional and (so far) non-deterministic 'hangs'[1] of > > subversion requests. It doesn't seem to matter if it's a commit or a > > status or a log. Eventually, at some point, a request to the server > > will hang. All requests are https. > > That's no good!!!
No it's not. It seems the problem (based on strace output) is related to apache connecting/reading to/from the ldap server. Every single hung apache instance was in a 'poll' on the ldap server socket. I still don't know why that is happening. We have other apache's doing the same thing and they don't exhibit this problem. > That...... is not good. I wonder if you do have an "I ran out of > randomness" issue? Can you roll back to the subversion-1.6.11 from > CentOs 5.6, or the subversion-1.6.15 from RPMforge (that I helped > bundle) on another server, and see if you can replicate the behavior? > Is this a wildly *busy* server? This is a wildly unbusy server. We host it on a dual proc Xen host and the load rarely gets above 0.1. -- seth /\ sethdaniel.org
