Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
I have 6 x Samsung SSD 840 pro drives in an RAID 0 configuration (mdraid). When I write to a non-DRBD partition on the drive, bypassing caches, I get 400 MB/sec. When I trim a filesystem mounted on a non-DRBD partition, it finishes fast. When I write to a DRBD replicated volume, I get 80MB/sec, which is about when I would expect using protocol C over a gigabit network. When I trim a filesystem that is mounted on a drbd device it is extremely slow. It takes three days to trim a 1.2TB volume. Running iperf between nodes shows 900Mbits/sec bandwidth, <1 ms latency, no packet loss. Why is trimming a DRBD volume so slow? This makes the servers unusable. -- Eric Robinson -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20170801/9331edb9/attachment.htm>