Hi!
How is the overall experience like re. deploying openser with mysql clusters? Are there gotchas etc that need to be taken care of? (For example, a 2006 article [1] says that "The MySQL NDB engine currently runs its database completely in memory. This means that you have to be able to fit your database in memory." But this is not documented as a limitation in mysql faq.) Non clustered experiments were occasionally catastrophic, so can't risk taking that route again.
Somebody suggested using drbd [2], but I did not quite like it -- it needs kernel modules, has more administrative overhead, is tied to linux, it is not clear how consistency (ie, state of in-memory data and on-disk data) is taken care of, and in any case ndbcluster appears to be the clustering solution with official stamp of approval.
(Related - is postgres any better? Apparently postgres' clustering capabilities are "better", but then cdrtool etc seems to be written with mysql as the primary target.)
Thanks, Sajith.
[1] http://www.oreillynet.com/pub/a/databases/2006/02/16/ha_mysql_cluster.html [2] http://www.drbd.org/
On Thursday 17 July 2008, Sajith T S wrote:
How is the overall experience like re. deploying openser with mysql clusters? Are there gotchas etc that need to be taken care of? (For example, a 2006 article [1] says that "The MySQL NDB engine currently runs its database completely in memory. This means that you have to be able to fit your database in memory." But this is not documented as a limitation in mysql faq.) Non clustered experiments were occasionally catastrophic, so can't risk taking that route again.
Somebody suggested using drbd [2], but I did not quite like it -- it needs kernel modules, has more administrative overhead, is tied to linux, it is not clear how consistency (ie, state of in-memory data and on-disk data) is taken care of, and in any case ndbcluster appears to be the clustering solution with official stamp of approval.
(Related - is postgres any better? Apparently postgres' clustering capabilities are "better", but then cdrtool etc seems to be written with mysql as the primary target.)
Hi Sajith,
recent MySQL cluster versions don't have this limitation (for non-index columns AFAIK) any more, you can store this data on the disk. There are some people that run openser with mysql cluster on this list, it was also discussed a few times in the past here.
I can't comment on drdb and on postgres clustering, but in openser is mysql the main target too. I did recently some work in this area, trying to even the differences between the database connectors. But mysql is IMHO still the most stable option, as it gets the most testing from the users.
Cheers,
Henning
Henning Westerholt henning.westerholt@1und1.de wrote:
recent MySQL cluster versions don't have this limitation (for non-index columns AFAIK) any more, you can store this data on the disk. There are some people that run openser with mysql cluster on this list, it was also discussed a few times in the past here.
Thank you Henning. I tried digging around a bit, but could not find what I was looking for. Sorry about the noise.
Thanks, Sajith.
Hi Sajith,
recent MySQL cluster versions don't have this limitation (for non-index columns AFAIK) any more, you can store this data on the disk. There are some people that run openser with mysql cluster on this list, it was also discussed a few times in the past here.
I think there is a penality about 10x when using disk-storage instead of memory-storage for ndb, mainly for write queries, so be carefull in your config file about the number of queries you make to the DB...
I can't comment on drdb and on postgres clustering, but in openser is mysql the main target too. I did recently some work in this area, trying to even the differences between the database connectors. But mysql is IMHO still the most stable option, as it gets the most testing from the users.
Cheers,
Henning
Users mailing list Users@lists.openser.org http://lists.openser.org/cgi-bin/mailman/listinfo/users
-
Henning Westerholt -
Sajith T S -
samuel