Hi there,
I did not think of answering right away, since I was only in the Cc:
list, but now it's been over a month and I have not seen any answer,
so here it goes anyway. :)
Since the e-mail concerns drbd mostly, I'll include Philipp Reisner
(drbd's author) in the Cc: list as well.
On Mon, Aug 21, 2000 at 09:14:38AM -0400, Pawul, Rudy wrote:
) Hello all.
)
) First, please excuse my writing in English, but I'm sadly not bi-lingual and
) thought the babel fish would make me sound like an idiot in Portugese....
Never mind, reading/writing/speaking in english is part of our daily
routine. We have people from other countries working here, and as we
also work with people from all over the world, english seems to be the
common communication protocol. :)
) Anyway, I'm about to try setting up a mail server cluster with Heartbeat and
) DRBD. I was hoping to get an idea of what I should expect - how stable it
) is, known problems, etc.
Both have been on use for production servers for quite some time,
without major problems. Bugs are usually found before they bite.
You should be aware though, that it is pretty easy to decrease overall
availability and consistency if you apply those tools in the wrong way.
) Alan had said that folks had "seen troubles" using drbd, but that connectiva
) was putting out solutions using it. I really couldn't get a feel of what to
) look for by reading the mailing list archives.
We have been working tightly with Philipp Reisner, drbd's author, for
quite some time now. drbd has had some troubles in the past, but those
show-stoppers have been squashed, and nowadays you can expect a good
level of stability from drbd for production servers.
) I was planning on using it on SuSE6.4, with ReiserFS, and the 0.4.8 rpm
) release of heartbeat. I guess I would grab the 0.5.6-pre1 drbd as well...
We have tested drbd and reiserfs some time ago, and we did find issues
that were solved. As of today you can use reiserfs on top of drbd,
even though a crash during drbd's resync phase (rejoin of a crashed
node) could leave reiserfs (or any other fs) broken. This is something
we have to live with, since drbd was not built to tolerate multiple
faults (and will never do, as long as it supports only two nodes).
I hope this helps. Further communication will be speedier. :)
Fábio
--
( Fábio Olivé Leite -- Conectiva HA Team -- olive@example.com )
( PPGC/UFRGS MSc candidate -- Advisor: Taisy Silva Weber )
( Linux - Distributed Systems - Fault Tolerance - Security - Pizza )
I did not think of answering right away, since I was only in the Cc:
list, but now it's been over a month and I have not seen any answer,
so here it goes anyway. :)
Since the e-mail concerns drbd mostly, I'll include Philipp Reisner
(drbd's author) in the Cc: list as well.
On Mon, Aug 21, 2000 at 09:14:38AM -0400, Pawul, Rudy wrote:
) Hello all.
)
) First, please excuse my writing in English, but I'm sadly not bi-lingual and
) thought the babel fish would make me sound like an idiot in Portugese....
Never mind, reading/writing/speaking in english is part of our daily
routine. We have people from other countries working here, and as we
also work with people from all over the world, english seems to be the
common communication protocol. :)
) Anyway, I'm about to try setting up a mail server cluster with Heartbeat and
) DRBD. I was hoping to get an idea of what I should expect - how stable it
) is, known problems, etc.
Both have been on use for production servers for quite some time,
without major problems. Bugs are usually found before they bite.
You should be aware though, that it is pretty easy to decrease overall
availability and consistency if you apply those tools in the wrong way.
) Alan had said that folks had "seen troubles" using drbd, but that connectiva
) was putting out solutions using it. I really couldn't get a feel of what to
) look for by reading the mailing list archives.
We have been working tightly with Philipp Reisner, drbd's author, for
quite some time now. drbd has had some troubles in the past, but those
show-stoppers have been squashed, and nowadays you can expect a good
level of stability from drbd for production servers.
) I was planning on using it on SuSE6.4, with ReiserFS, and the 0.4.8 rpm
) release of heartbeat. I guess I would grab the 0.5.6-pre1 drbd as well...
We have tested drbd and reiserfs some time ago, and we did find issues
that were solved. As of today you can use reiserfs on top of drbd,
even though a crash during drbd's resync phase (rejoin of a crashed
node) could leave reiserfs (or any other fs) broken. This is something
we have to live with, since drbd was not built to tolerate multiple
faults (and will never do, as long as it supports only two nodes).
I hope this helps. Further communication will be speedier. :)
Fábio
--
( Fábio Olivé Leite -- Conectiva HA Team -- olive@example.com )
( PPGC/UFRGS MSc candidate -- Advisor: Taisy Silva Weber )
( Linux - Distributed Systems - Fault Tolerance - Security - Pizza )