[DRBD-user] Sync target performance issue

Rafe Slattery Rafe.Slattery at iquate.com
Thu Nov 23 11:31:38 CET 2006

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



I'm working on a drbd installation which is having some performance
issues that I was hoping I could get some help with

The two servers involved are hp proliant dl 380's with lvm storage. The
controllers are Smart Array 6i 's and the nics are some broadcom Gb
Ethernet jobbies. They are linked by a Gb switch.


I was initially running on v7.X with kernel 2.6.9 but this resulted in
sync rates of ~1MB/sec.

To try to sort this I updated to v8pre6 and kernel 2.6.18 and added in
Netgear Gb Ethernet PCI cards. This combo got me up to about 7MB / sec.

But when testing this yesterday while running a full sync, the
performance on the sync target was abysmal, to the extent that heartbeat
was unable to operate. Also on the drbd primary node, reads were fine
and dandy but anything which involved writing to the shared partition
was abysmally slow (30 secs to write a 2k file from vi).


Any suggestions would be greatly appreciated.







global {

    minor-count 5;

    dialog-refresh 5; # 5 seconds




resource db {

protocol C;

 startup {

   # wfc-timeout  0;

   degr-wfc-timeout 120;    # 2 minutes.


 syncer {

   rate 700000K;

   al-extents 257;




 max-buffers 4096;

 max-epoch-size 1024;



 on monitor.localdomain {

   device     /dev/drbd0;

   disk      /dev/VolGroup00/LogVol02;


 #  address;

   meta-disk  internal;



 on monitor1.localdomain {

   device    /dev/drbd0;

   disk      /dev/VolGroup00/LogVol02;



   meta-disk internal;




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20061123/d6dc0476/attachment.htm>

More information about the drbd-user mailing list