Exchange migrations, scenarios to hope you're never in.



  • I wanted to add a public note about what happened with a project this past weekend. The first piece of this project we're tackling is getting Exchange migrated from a SBS2008 instance to a 2016 VM (management doesn't want to use online services, local IT person doesn't like the decision. So we get to go through this pain.)

    So we know that you can't go directly from Exchange 2007 to Exchange 2016. A migration to Exchange 2013 is required to happen in between. As it turns out, the Exchange 2007 to 2013 migration process eats massive amounts of drive space. The current 2007 database store is rather large (another battle the local IT person can't win, users cleaning up their inboxes), but nothing extreme.

    The 2013 database when we were actually doing the batch migration of mailboxes grew to 140GB with another ~200GB of log files. It ran itself out of drive space twice. Thankfully this is at least a proper setup with Hyper-V and we were able to increase the drive space twice. The first time it picked right back up and continued on after getting the drive space increased. The second time however, the main database had been corrupted. We ended up being able to fix the database corruption using eseutil. We also had to move all the log files out of the directory before Exchange would remount the database, which at least makes sense.

    Just wanted to post a public note because this issue doesn't appear much online, and it would have been really good to know about this "bug" before this actually happened.



  • User abuse is the only bug we can't fix with technology.

    Good to know eseutil was able to correct the issue.



  • @dustinb3403 said in Exchange migrations, scenarios to hope you're never in.:

    User abuse is the only bug we can't fix with technology.

    Good to know eseutil was able to correct the issue.

    Even that was just very weird. The status had stopped updating and processes associated with it quit doing anything for around 20 minutes. We were both wondering what was going on when I was like "I'm going to ctrl+c it." So the cancel current command actually kicked off a whole bunch of activity that went quite quickly compared to what it had been running at before and actually repaired the DB. We were both more than a little confused over that one.