large volume of failed sets and deletes

[Gil Hildebrand Jr]

I setup memcached fairly recently and have noticed a large number of failed
set() and delete() requests, on the order of 10-20 per min. I have 4 1.5 GB
servers running on 2 hosts, and am using the MemcachedAggregator PHP
wrapper.

PHP Notice:  Memcache::set() [<a href='function.set'>function.set</a>]:
marked server '172.28.0.22:11211' as failed

I do not see this notice for get(), only set() and delete(). When one of
these requests fails it can sometimes cause stale data to enter the cache.

Here is a stats dump from one of my instances (all have roughly the same
stats):

stats
STAT pid 5078
STAT uptime 417260
STAT time 1181066104
STAT version 1.2.1
STAT pointer_size 64
STAT rusage_user 806.357415
STAT rusage_system 2371.098538
STAT curr_items 454841
STAT total_items 2729384
STAT bytes 1396823505
STAT curr_connections 17
STAT total_connections 7924721
STAT connection_structures 482
STAT cmd_get 26983062
STAT cmd_set 2729384
STAT get_hits 22592930
STAT get_misses 4390132
STAT bytes_read 67371856639
STAT bytes_written 37621419623
STAT limit_maxbytes 1610612736
END

Any idea what might be going wrong? Should I switch to the PECL extension
instead of the wrapper class?

Thanks,
Gil
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.danga.com/pipermail/memcached/attachments/20070605/0d9a98fd/attachment.html