> That isn't a terrifically high load. Can you do: 
>
> `ls -l /proc/$(pidof memcached)/fd | wc -l` ? just to confirm you aren't 
> hitting maxconns in a weird way (as netstat doesn't show a ton of stuff 
> open). 
>
> Do you tend to run your other servers at or near maxconns? 
>
> Your box isn't out of TIME_WAIT buckets or ephemeral ports, or netfilter 
> maxed out? (check dmesg), or some other issue that would prevent it from 
> getting new connections until you restart something?
>  
>
 
server too far for maxcon or fds limit
 
ls -l /proc/$(pidof memcached)/fd | wc -l
~1.5K on each server
 
netstat -tpn|grep TIME_WAIT|wc -l
~100
 
netstat -tpn|grep ':11211'|wc -l
1454
 
cat /proc/net/sockstat
sockets: used 1642
TCP: inuse 1544 orphan 0 tw 91 alloc 1547 mem 386
UDP: inuse 5 mem 1
UDPLITE: inuse 0
RAW: inuse 0
FRAG: inuse 0 memory 0
tcp_fin_timeout 5
net.core.somaxconn 512K
net.ipv4.tcp_tw_reuse 1
tcp_max_tw_buckets 2M
tcp_tw_recycle 1
backlog  30k
net.ipv4.ip_local_port_range = 1024     65535
 
server running memcached only, no other services
 
conntrack is empty
iptables\netfilter not in use
 
logs clean
 
memcached user limits like nofile & nproc is set to 1M
 
memcached working fine few days with the same load ops\req/sec.
then memcached hangs
 
what additional info (trace\debug) can i provide if memcached will hang in 
next time?
 
Thank you

-- 

--- 
You received this message because you are subscribed to the Google Groups 
"memcached" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.


Reply via email to