Anyone else running into connection latency on 8.1.18? We were at 8.1.17.008 and 8.1.17.100, and came to 8.1.18.0 due to some rollup locking patches (vs going to 8.1.17.015).
Now, we have pretty substantial hangs for client and dsmadmc connections. During our normal backup window, it gets bad enough that SSL initialization fails, and we end up with a lot of TDP failures (especially TDPO because of how many sessions they create). All the way back at 8.1.6 in 2018, we bumped our maxsessions to 1200 because we'd bump up against 800. We've been okay since then, and until the 8.1.18 update. Now, at around 400 sessions, a new dsmadmc session from localhost can take 2-3 minutes to prompt for password. Verified it's not a DNS timeout. Turned off DNSLOOKUP and no change. Not a server specific thing. We have a large number of AIX servers in more than one datacenter, and we have the same issue on both. We have the same issue connecting to localhost. Servers with fewer clients are better off. AIX 7.2.5.5, 250GB of RAM, 20x POWER8 cores, full system partition. CPU peak is 45-55%, and the system is very responsive. Peak network is now around 550MB/sec. I'm pretty sure we were more than double that before. Lots of 10gbit ports. Only 2% RAM free, but no paging activity, and numclient is still 12% with a good number of LRUable pages. No ssh latency. No lag at the db2 layer. DB is 3.6TB, but it's heavily fragmented, and probably more like 2.8TB used if we did an offline reorg. No DBB, expire, reclaim, etc. running during the backup window. I do see an automatic runstats on SD_CHUNK_LOCATIONS. Anyway, we have a case with IBM, and are slogging through different trace attempts. It's just been an issue long enough, I was hoping to get some commiseration if anyone else has noticed anything similar. With friendly Regards, Josh-Daniel S. Davis