Hi!
> Either "pull-ahead" doesn't work, or another configuration interferes with
> this setting,
Finally I found the reason why it didn't work as expected. It was a small
sentence in this description:
https://www.drbd.org/en/doc/users-guide-84/s-configure-congestion-policy
-> A good value for congestion-fill is 90% of the TCP network send buffer, in
non-DRBD Proxy setups.
My former "congestion-fill 2G" and "sndbuf-size 512k" could never trigger the
"pull-ahead" state, because the configured fill level simply was never reached.
This is now a working configuration:
disk {
al-extents 2000;
# Full speed for 1GBit link
c-plan-ahead 0;
resync-rate 110M;
}
net {
max-buffers 8000;
max-epoch-size 8000;
sndbuf-size 512k;
cram-hmac-alg sha1;
shared-secret "xxxx";
verify-alg md5;
protocol A;
on-congestion pull-ahead;
# should be 90% of the TCP send buffer if no DRBD Proxy is used
congestion-fill 460k;
# should be 90% of the al-extents
congestion-extents 1800;
}
So again: Good job, Linbit!
Maybe a sample configuration in the documentation for such scenario would be
nice for others.
BR,
Jasmin
_______________________________________________
drbd-user mailing list
[email protected]
http://lists.linbit.com/mailman/listinfo/drbd-user
