Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Thu, Feb 12, 2009 at 05:02:32PM +0100, Julien Reveillet wrote: > Hello, > > I have a very special configuration to set up. > > I need to have 2 servers running CentOs 5 with a mysql server 4.0.21 ( > that a must do...). > The data base directory needs to be replicated between both servers with > one mysql server running with read-write access and the other with > read-only access. > > The production server (mysqld with rw access) which is highly loaded > needs to not be impacted by drbd accesses ( it can be a little bit ) and > the web server (mysqld with ro access) can be asynchronous for some > minutes with no problems. > > For now, i choosed to use drbd 8.2.6 with ocfs2. > It is set up and running correctly on the same lan with a descent sync > rate but i will have to try it soon over a wan network. > > On what i've tested and read, i can't use protocol A for this scenario > as the option "allow-two-primaries" needs only protocol C. > > The whole database directory is about 70GB which is quite huge to full > sync over a 2M wan network. For sure, i will do the first sync over a > lan but i'm afraid of cases where drbd will go to standalone and will > need a full resync. > > So my questions are : > > - Falling to standalone mode will automaticly needs a full resync? No. Why? > - Can this kind of scenario be possible to work on or not ? > (if not, do you think of another solution?) If you _want_ asynchronous replication, why would you go for a synchronous replication solution? why not go for mysql replication, if that is actually what you want? cluster file lock manager latency and all sorts of issues with a WAN link will be a real pita. > - Will protocol C cause many heavy lags on my production server ? > (i guess yes but...) the replication link is 2 megabit/second, which is roughly 200 kByte per second. and the latency is what? 200 msec? well, DRBD cannot magically improve those numbers. I don't think that would be feasibly for a cluster file system. and not at all for a data base on top of a cluster file system. -- : Lars Ellenberg : LINBIT | Your Way to High Availability : DRBD/HA support and consulting http://www.linbit.com DRBD® and LINBIT® are registered trademarks of LINBIT, Austria. __ please don't Cc me, but send to list -- I'm subscribed