Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Lars Ellenberg <Lars.Ellenberg at linbit.com> on 04/12/2004 08:19:17 AM
To: Jesse A Dubin <jdubin1 at lsu.edu>
cc: drbd-user at nudlaug.linbit.com
Subject: Re: [DRBD-user] DRBD sync rate way low
DRBD knows nothing about your file system, it is a block device,
below the file system level.
With DRBD 0.6.X, whenever a Primary goes down, there has to be a
full sync. Even if *you* know that nothing, or only "little", was
changed, DRBD has no way to know that. It only knows that the
primary vanished unexpectedly, and concludes it needs to sync...
----------------------------------------
I see.....
----------------------------------------
> ....I think this would explain
> why my data is missing and my server is down....
I still don't understand what you mean with "my data is missing"...
Please give more detail about
what you are doing, and
what you expect to happen, and
where "real world" does not meet your expectations.
----------------------------------------
Well, before the reboot, I had data in my node dirs. After the reboot,
this information is gone, as well as the whole directory structure. If it
were just a DRBD issue, as you said, I would still have data on the primary
node and I would still be serving pages.
I expected that upon the shutdown of the primary, there would be a failover
and the Secondary would promote itself (it started to, but I think the
Primary came back up before the Secondary could take over all but one node)
and the Primary, upon its reboot, would become Secondary.
This seems to have happened as far as DRBD is concerned, but there's no
data in the dirs being replicated anymore.
I'm running DRBD 0.6.10 on SuSE 8.2. drbd.conf (minus comments) looks
like:
resource www {
protocol = B
load-only
disk
do-panic
disk-size = 8388608k
}
net {
sync-max = 100M # if you don't care about network saturation
# -max has to be larger than -min, obviously
sync-min = 500k
# sync-max = 1M # maximal average syncer bandwidth
tl-size = 5000 # transfer log size, ensures strict write ordering
timeout = 60 # 0.1 seconds
connect-int = 10 # seconds
ping-int = 10 # seconds
}
on indigo {
device = /dev/nb9
disk = /dev/vgr5/www
address = 192.168.101.101
port = 7797
}
on origin {
device = /dev/nb9
disk = /dev/vgr5/www
address = 192.168.101.102
port = 7797
}
}
This is mostly identical for each of our nodes.
DRBD has been running on our servers for a while now, and has generally
been working fine and has recently been handed down to me.
----------------------------------------
Thanks.
Jesse
-----