shibd crashes if memcache server is marked dead

Description

If a memcache server is marked dead and kicked out of the list, shibd goes into a recursive loop and crashes the daemon.

Environment

None

Attachments

2
  • 10 Sep 2009, 05:35 AM
  • 19 Aug 2009, 12:12 PM

Activity

Scott Cantor October 28, 2009 at 2:07 PM

http://svn.middleware.georgetown.edu/view/cpp-sp?view=rev&revision=3148

Patches applied with two compile fixes. I noted a change to the processing done to save off the memcache hosts to use, I assume that's been vetted and is appropriate.

Cal Heldenbrand September 10, 2009 at 8:49 AM

Awesome, thanks Andre!

André Cruz September 10, 2009 at 5:35 AM

Here is the patch with the new configuration options:

Ex:

<StorageService type="MEMCACHE" id="mc" prefix="vm-sso-tst1.vm2.bk.sapo.pt:" sendTimeout="1000001" recvTimeout="1000002" pollTimeout="1003" failLimit="6" retryTimeout="31">
<Hosts>
127.0.0.1:11211
</Hosts>
</StorageService>

Cal Heldenbrand August 20, 2009 at 2:32 PM

Also if Andre has the time to code this, it would be nice to have user configurable timeout settings. I have the plugin compiled with the limemcached default settings right now (-1 timeout, blocking) and it seems to be more reliable for me than the 1 second timeout.

Thanks Andre!

--Cal

Cal Heldenbrand August 20, 2009 at 2:28 PM

This appears to be working fine, so please include this patch in the next release.

Thanks!

--Cal

Fixed

Details

Assignee

Reporter

Fix versions

Affects versions

Created August 19, 2009 at 12:12 PM
Updated November 4, 2009 at 10:22 AM
Resolved October 28, 2009 at 2:07 PM