[DRBD-user] degraded write performace

Duane Cox duanec at mail.illicom.net
Thu Mar 29 22:44:50 CEST 2007

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


To recap the benchmark tests...

two identical servers (hardware and software)
/dev/sdb is a single 36G 15K U160 scsi seagate
kernel-2.6.20.3 with drbd-8.0.1

[drbd module NOT loaded]
./dm -a 0 -b 1M -s 3G -y -m -p -o /dev/sdb ... [ multiple concurrent tests
yield ~70 MB/sec ]

[modprobe drbd (no ps running, not started, just loaded)]
./dm -a 0 -b 1M -s 3G -y -m -p -o /dev/sdb ... [ multiple concurrent tests
yield ~70 MB/sec ]

so far, so good....

[drbdadm adjust all] (worker thread started) (drbd in "st:Secondary/Unknown"
"ds:UpToDate/DUnknown" state)
./dm -a 0 -b 1M -s 3G -y -m -p -o /dev/sdb ... [ multiple concurrent tests
yeild ... ~70 MB/sec initially and degrades with each additional passing of
the benchmark.
After 2 tests, performance is down to 50 - 60 MB/sec.
After 10 tests performance is down to < 30 MB/sec.
After 20 tests performance is down to < 25 MB/sec.

[drbdadm disconnnect all] (acutally not needed, in disconnected state
already)
No change, performance has not changed.  Still degraded to ~ 25 MB/sec.

[drbdadm detach all] (worker thread killed) (module still loaded)
BIG change!!!  Benchmark now up to ~70 MB/sec !!!

Tests conduced on two identical machines, both provided identical
back-to-back results.





More information about the drbd-user mailing list