Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On 09/23/2010 02:52 AM, Alex Adriaanse wrote: > These hangs seemed to coincide with times when there was a large spike > in memory consumption and much of the server's physical memory was used > up, resulting in a significant increase in swapping. The server has 4GB > of RAM and 3GB of swap space. The most swap space I've seen in use was > 0.9GB (during the periods of heavy memory consumption). However, I > wasn't able to measure actual swap usage during these freezes, so I > can't confirm this correlation or the exact swap usage during the > freezes. I have seen this problem as well. You could try, as a workaround, to add "data-integrity-alg crc32c;" to the net section. This resets the connection if an error is detected and reconnects again. Since I have done that, the timeouts are gone. But I see several times a day "Digest integrity check FAILED" messages. This seems to happen at high IO and CPU load. I posted a message before on this list on how I can reproduce. Just run "pi 22" with a /tmp partition that is backed by drbd (that's not your setup as I read). > Some background information: the /usr, /var, /var/log, /home, and /srv > filesystems run off various DRBD devices, which use LVM logical volumes > as the underlying storage, which in turn uses two hard drives mirrored > using MD RAID1 as its physical volumes. The DRBD devices are configured > as Primary role, with the Secondary server being connected over a long > distance link. The root filesystem, /tmp, and swap bypass DRBD and use > LVM logical volume directly. These logical volumes reside on the same > physical volume as the logical volumes that back the DRBD devices > mentioned above. This looks like a complicated setup. Is this because you cannot run the entire machine from a drbd disk? Is this intended for failover? Theo