I am using memcached on the same server (Ubuntu 9.04 Jaunty 32-bit on
EC2) as a Merb app (relatively low traffic), using memcache-client
(pure ruby) on the client-side.
About once or twice a day we are seeing ECONNREFUSED when the client
is attempting a get.
A useful clue is that when it happens, it happens repeatedly for the
next 10-20 seconds.

I have tried the following and am at a loss of what to do next.

Looked at memcached stats.
STAT pid 15943
STAT uptime 330202
STAT time 1260309445
STAT version 1.4.3
STAT pointer_size 32
STAT rusage_user 0.332020
STAT rusage_system 0.828051
STAT curr_connections 12
STAT total_connections 36
STAT connection_structures 13
STAT cmd_get 30961
STAT cmd_set 1243
STAT cmd_flush 0
STAT get_hits 28561
STAT get_misses 2400
STAT delete_misses 0
STAT delete_hits 780
STAT incr_misses 0
STAT incr_hits 0
STAT decr_misses 0
STAT decr_hits 0
STAT cas_misses 0
STAT cas_hits 0
STAT cas_badval 0
STAT bytes_read 5944099
STAT bytes_written 63738254
STAT limit_maxbytes 67108864
STAT accepting_conns 1
STAT listen_disabled_num 0
STAT threads 4
STAT conn_yields 0
STAT bytes 1266505
STAT curr_items 461
STAT total_items 1243
STAT evictions 0

Nothing stands out here, to my relatively untrained eyes.
Listen_disabled_num is 0.
We have 4 apache processes.

Also tried turning on tcp_tw recycling and reuse.

I tried opening up the site in a bunch of tabs to see if I can trigger
it but have not yet been able to (that's probably my browser
throttling the number of connections?)

Didn't see any red flags in netstats either.

Any help on where to dive into next is appreciated.  (All the above is
what I could extract from the memcached faq and casual googling.)

Thanks,
Gary

Reply via email to