not sure if it relates to LXD or not, so I’m seeking for guidance… Here is my case: I’ve built a server based on RaptorCS Blackbird motherboard, with a 8 cores Power9 processor and 128MB Ram. I intend to use it mainly as a Postgresql Server for Datawarehousing.
Now I made some pgbench tests, both directly on the main system, and within a LXD container.
Main system uses a single Samsung 840 PRO SSD (sda), while LXD data pool is a ZFS pool with 3 Samsung SSD (sdb, sdc, sdd1 and sdd2).
Using the same pgbench (-scale 10000 -Time 600) and the same postgresql.conf tuning, I get a quite satisfying 2000TPS on the main system, but only 113 TPS on the LXD instance,with a bad latency average of 263.931 ms !..
Looking at the logs, I found this (maybe) suspicious line:
Sep 04 13:05:30 black-pg iprinit: Timeout reached. Ensure the sg module is loaded, then run iprinit manually to ensure all ipr RAID adapters are running optimally
Has anyone experienced such behaviour, and how would you troubleshoot this ?
Thanks in advance,