According to the Pervasive site and docs an Active-Passive (only?) MS
Cluster works and is supported. Nothing on 3-node (or more) clusters there.

It took some creative configuration, but I managed to build an
Active-Active-Passive cluster. Fail-over and all working fine.

We would like to use this for a customer, but does anybody know if this is
supported by Pervasive? Anyone had actual experience with 3-node (or more)
Pervasive clusters?

It is probably not supported by Pervasive but does depend on specific
configuration.

The Pervasive.SQL database engine does not know how to "share" data
files between multiple engines, so multiple database engines can not
access the same data file at the same time.

If the different nodes are supporting different databases / sets of
data then it may be ok, but it brings up the transaction logs the
engine uses. Again the transaction logs are exclusive to a single
engine. The transaction log file location is configurable and if it
is put on the same volume as the data that would be fine, but, if both
nodes failed over at the same time, there is only one transaction log
location so you would still be out of luck.

If you turn off transaction logging and each node is working on its
own data files then you should be fine, otherwise I would expect to
have problems.

Loenard

On Wed, 19 Apr 2006 14:16:24 +0200, "Bob"
<google--no-spam--@...ad.heuvel.nu> wrote:

The cluster contains 2 resource groups, each containing (besides Pervasive,
ip, share, name) it's own shared disk on which both the data files and
transactionlog directory is placed. So both running instances, on different
servers, use a different drive (letter) for the transactionlog directory.

Assumption was that it will not be possible to have 2 instances of Pervasive
running on one node at the same time (partly because of the single
configurable location for the transaction logs). So this 3-node cluster will
only provide full (Pervasive) access, as long as no more than 1 node is
down.

So I have this (test(ed)) environment up-and-running, but are wondering if
anyone has done this before, in a real production environment, and if
Pervasive knows this is possible (since I can't find any mention at all on
3-node, or more, clustering of Pervasive).

"Leonard" <lharvey (AT) austin (DOT) rr.com> wrote

When you say two resource groups -- does this imply two different
applications? If so, then this should be OK, since each system will
look like Active-Passive anyway.

I helped a customer set up a similar configuration with NetWare
clustering, whereby they had 3 servers with two of them hosting active
Pervasive databases, and any one could fail over to the standby server.

What was cool was that BOTH could fail over to a single box, as well.
This was possible because they were more worried about downtime and
performance, and less worried about perfect data -- so they disabled
Transaction Logging. This eliminated any problems from the TxLogs
during failover.

If you have a similar config -- two completely independent databases,
then you MAY be able to get a double-fault to work -- but it might get
tricky with the translog data. You might be able to start the engine,
let the logs roll forward if needed, then reconfigure the system for
the second directory and do the same there. It won't be fool-proof,
and it'll take some manual work, but it just might work out for you...
Goldstar Software Inc.
Pervasive-based Products, Training & Services
Bill Bach
BillBach (AT) goldstarsoftware (DOT) com
http://www.goldstarsoftware.com
*** Austin: Pervasive Service & Support Class - 05/2006 ***
*** Chicago: Pervasive Service & Support Class - 07/2006 ***


Bob wrote: