Hi @all,
after few weeks and months i want to recapitulate my experience with
Sequoia and wants to let you know some bugs and some improvements which
maybe can release in the next versions of sequoia.
First of all we use Sequoia 3 Beta 2.
Our environment (four server|RAIDdb1):
RZ 1 => Controller 1 , DB-Node 1
RZ 2 => Controller 2 , DB-Node 2
Sequoia is a really great opensource project and we enjoy the time of
implementing the software etc. We love the stable operating and the
simpling administration of this software. But after a few situations of
testing we noticed that some things can confuse the total cluster
environment. For example, if we connect to the cluter using the
Clustersoftware "DB - Visualizer" and send a wrong statement like:
SELECT * FROM scheme.tab; commit
this statement will stay in the queue and you can't delete it from the
queue (i know in Beta 3 "Killer"-command will exist for something like
this). The problem is, that you can't create backup of the DB because
this statement is still in the queue.
Other programs like SquirrelSQL or iSQL send an error message to the
user and don't send this statement to the cluster. So we are not allowed
to use DB-Visualizer anymore ;)
We used to send this statement directly to the Database and postgres
thrown an exception like:
ERROR: Error »Syntaxerror« at »SELECT«
ROW 2: SELECT * FROM scheme.table;
So we think that sequoia can't handle this error-message. Maybe a JIRA
have to be open up for this problem.
Another solution could be a "statement-timeout". If a statement like
that one above is staying in the queue, after ~60 seconds it will be
deleted because of inactivity ...or something like this ;)
Another thing is our group communication using appia. If the Debugmode
for hedera is on, everything works fine. But if we configure it to info
or warn the group communcation doesn't work...quite weird. We didn't
find a solution yet...our infrastructure-section said that everything is
ok on our ciscos to use multicast....mmh, i'll compile Beta 3 from cvs
in the next time to see if something like this will happen to with the
new version.
It would be cool if a timestamp could be implemented to console. So i
can choose names for our dumps like dumpname_timestamp...for example...
Is this possible?
combined i only can say that sequoia is one of the best clustersolutions
i know, thx for the development :)
Greets
Chris
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
