A bit of background - recently we've been upgrading Win2003 systems to Win2008, and all of these servers have an LSI RAID controller card installed. The previous system image had popup and email alerts disabled in the MRMonitor config, but someone overlooked these checkboxes in the 2008 installation image. We've already raised the flag on that problem.
However, I've now two systems that suffered a drive failure since the upgrade, and while there is no longer any issues with the RAID since replacing and resyncing the drives, at every login, we get a stack of desktop popups advising of write failure!
I've verified that the problem is not ongoing (there are no new entries in either the system event logs or the MRMonitor logs), and I've also gone into MRMonitor's configuration, cleared the checkboxes for popup and email alerts, and sent the config back to the LSI card. I've confirmed (by explicitly loading the config from the card) that these checkboxes are indeed cleared.
Yet upon login, unless I disable the MRMonitor service outright, I get about 20-30 popups, all advising of a write error to sectors of the drive. It's not limited to my account, either - I've received complaints from others logging into the same systems on their accounts.
I've saved/cleared the MRMonitor log, archived the system logs, and both restarted and cold-booted the system. Google searches refer to LSI MegaRAID being run on Solaris (these are indeed SUN Xfire systems, but they are running Windows) so that avenue is sketchy at best.
Does anyone have an idea of where these event notifications are cached? I'd hate to have to reimage entire servers just because of a noisy situation with MRMonitor.
Thanks very much.
The remedy to this issue turned out to be the stock remedy for nearly all ill-behaved software under Windows.
Uninstall MegaRAID 2.90-01
Reboot.
Reinstall MegaRAID 2.90-01.
Reconfigure alerts within MRMonitor from stock config so that all alerts are sent to logs only, no popups, no emails.
Since this was originally deployed as an image file, uninstalling/reinstalling individual components on a server by server basis is not an efficient method, BUT... one-offs are not an issue. We can handle that.