For cost reasons I
have suggested for our company to use LCR as a high availability solution for
our upcoming Exchange 2007 server. So I have requested to get 2 extra sets of
hard drive spindles. One for the database
copy and another one for the logs copy. A coworker disagrees with me because he
says that my solution is what he calls a “Poor man’s SAN” . He says that there
is an existing SAN in the company with the capability of creating snapshots of
the databases and the logs as often as we want. So if the database become
corrupt we would revert to the latest snapshot and replay the logs generated
after the database became corrupt. I partially agreed and told him I was going
to think about it.
Here I am thinking
about and seeking opinions and clarifications to possible misunderstandings on
my part on how the Exchange recovery of databases works. Here are some points I
need to have clear:
1-
How
LCR works?: Once a database becomes corrupted I would manually point the
settings to look for the copy of the database and the copy of the logs in the
volumes where they reside. No replay of logs that I need to initiate. Is this
correct?. Few minutes to get this done
2-
Replay
of logs: If the SAN solution is used, wouldn’t I have to use ESEUTIL to replay
those logs and somehow determine which logs were generated after the original
database crashed? . This might take more than few minutes depending on the size
of the database. Is there any other way to replay those logs?
Please help me put a
perspective on this. If I am correct on my assumptions, SAN snapshots solution
does not seem to be the best option between the 2 choices. With LCR, we could
probably still use the SAN adding 2 extra volumes for a copy of the database and
the logs.
What do you think?