Originally posted by: SystemAdmin

I have an HACMP cluster and have noticed that I have UDP packets being dropped due to no socket. It tends to be around 32 or so a minute and I wanted to know if there is something I should be tuning to correct these. All the documentation says to investigate the application - but I think the application is HA as I don't see numbers this high on my other busy non_HA systems. I am not seeing any UDP socket buffer overflows.

Any help would be appreciated. Thanks.

Sockthresh is defaulting to 85% and my no UDP settings are:
udp_bad_port_limit = 0
udp_ephemeral_high = 65535
udp_ephemeral_low = 32768
udp_inpcb_hashtab_siz = 24499
udp_pmtu_discover = 1
udp_recvspace = 65536
udp_sendspace = 32768

rfc1323=0 but we plan to change that to 1 shortly and increase our send and recv space for both UDP and TCP.
#AIX-Forum

Originally posted by: mahespth

UDP packets being dropped due to no socket is usually due to high NFS traffic;

"NFS packets are usually dropped at the socket buffer only when a server has a lot of NFS write traffic. The NFS server uses a UDP socket attached to NFS port 2049 and all incoming data is buffered on that UDP port. The default size of this buffer is 60,000 bytes. You can divide that number by the size of the default NFS Version 3 write packet (32786) to find that it will take 19 simultaneous write packets to overflow that buffer."

You can check "nfso -a" and "nfsstat" if you think you are a heavy NFS user. If not then look at udp_(send/recv)space and sb_max.

Regards
Steve Maher
#AIX-Forum