Tomasz Chmielewski wrote:
> I have "IPVS: ip_vs_send_async error" literally flooding in dmesg.
This usually happens when there are too many connection updates to
send in the default time (was around half a second IIRC) and they
don't fit in the send buffer.
That's probably not the case right now though.
> I had it in the past, but somehow solved it with this in sysctl.conf:
> net.core.wmem_max = 16777216
> net.core.rmem_max = 16777216
> Now it happens again, with barely no traffic (40 connections in DR mode
> or so).
> What should I do to fix it?
Can you check with tcpdump if there are connection sync packets being
sent on the wire and how often/how long they are (host 18.104.22.168 and
udp port 8848)?
Run netstat -npua (or similar) to see if there are listening sockets
on 22.214.171.124:8848 or any sockets that connect to 126.96.36.199:8848?
If there is a missing socket and no packets, perhaps something
happened with the sending socket and it got closed (would it close if
it got ICMP unreachable in response)? There does not seem to be any
more error handling than this message - so if the socket indeed got
closed (somehow) then it would never be detected and repaired and you
would just get one of these message at each sync sending attempt.
Please read the documentation before posting - it's available at:
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users