I've had this happen when memcached segfaulted partway through operation ...
there may be other ways to reproduce this fault.
Basically, the PHP client goes into an infinite loop trying to get a valid
socket to a server. Because no socket is available, the loop never exits (until
it exceeds the PHP execution time limit).
I've attached a patch to fix this.
Steps to Reproduce:
Created attachment 15096 [details, diff]
Patch to fix infinite loop problem
Created attachment 15101 [details, diff]
Patch with fix for _set() robustness problem
This patch has the inifinite loop fix, and also traps a problem where the
expiry time would not be passed across the network to memcached.
With these patches, the PHP client is now working for me on a production
machine (so far ;-)
Created attachment 15113 [details, diff]
Latest diff for improved robustness
This latest patch further improves the robustness of the PHP API when memcached
I'm going to incorporate these diffs into the ebuild while waiting for a new release.
Look for -r1 to be out with some cheap hack to get around start-stop-daemon (init script) and these patches some time Tuesday.
gonna close this since 1.1.7 is out and the php api was retouched
Created attachment 101892 [details, diff]
Created attachment 101893 [details]