We recently upgraded from SmarterMail Enterprise Build 7503 to Build 7523 using the .EXE on Windows Server 2019 on Hyper-V. We started getting @ 6 crashes per day that are referencing .NET clr.dll with the error:
The process was terminated due to an internal error in the .NET Runtime at IP 00007FFF715A59C3 (00007FFF71550000) with exit code 80131506.
We seemingly resolved these by discovering and eliminating the following issues:
- Using Dynamic Memory on the VM in Hyper-V. This caused an apparent memory leak that would otherwise use up 100% Memory resulting in the mailservice.exe to crash > 6x/day with the error above. Disabling this setting in the Hyper-V Host eliminated this issue and vastly decreased the memory usage. This probably should be isolated and fixed in SmarterMail at some point but I'm willing to accept that's more an environmental issue than a SmarterMail issue. Build 7503 would periodically crash < 1/day with the error message above, but it did not have the memory leak.
- We had multiple IMAP users with > 500 simultaneous IMAP sessions each. Each of them were MacOS Mail users. They would only utilize @ 10-16 simultaneous IMAP sessions on Build 7503. Turns out each of these users had a folder at the root level (on the same level as their normal Inbox, Deleted Items, Drafts, Sent Items) that was named with their email address. This folder then contained additional Archive, Inbox, Deleted Items, Drafts, Sent Items sub-folders (so \email@example.com\Archive, \firstname.lastname@example.org\Inbox, \email@example.com\Deleted Items, etc.) This was causing MacOS Mail IMAP connections to go crazy. These folders existed for these users in Build 7503 and did not cause this problem. Removing these folders named after their email address, and all their sub-folders, immediately resolved the high number of simultaneous IMAP sessions for those users in Build 7523.
- We had numerous users with sub-folders containing messages located in the root Deleted Items folder. Again, each of these users were MacOS Mail users and Build 7503 didn't have an issue with these sub-folders in the Deleted Items folder. However, with Build 7523 it was all of a sudden a problem and leading to Indexing issues and high memory usage. Deleting these sub-folders out of the Deleted Items folder immediately resolved the problem on Build 7523 and Indexing was able to complete for those users and memory usage was dramatically reduced. It also dramatically reduced the number of simultaneous IMAP sessions for each user just as the issue in #2.
- We had a closed Mailing List that was a recipient on another closed Mailing List in SmarterMail. This is a fringe case that never should happen but it did (and the user that added them has been scolded). The two of them kept bouncing "You do not have permissions to post to this list" messages back and forth until we had 100,000+ messages backed up in the Spool. Again, Build 7503 handled this fine (it stopped bouncing after a half dozen bounces back and forth) but on Build 7523 it will go on forever until manually resolved by an admin.
- A user with an autoresponder enabled in their Outlook CC'd themselves on an Outgoing message. They had themselves set as a Trusted Sender in SmarterMail. This resulted in an endless loop of emails from himself to himself until we had 30,000+ messages backed up in the Spool. This confirmed that Message Throttling is not working in Build 7523, which I suspect may have been the issue with #4.
Once these 5 issues were remedied Build 7523 has been stable for 2 days and counting without a single crash referencing .NET clr.dll. Even though these are all either environmental or data induced, items 2 & 3 definitely are due to a bug in SmarterMail and 4 & 5 due to a recent change.
We also made the following discoveries of issues with Build 7523:
- Message Throttling for messages, bandwidth and bounces is no longer working (can confirm that it was as of Build 7503).
- Autoresponders no longer function (can confirm that it was as of Build 7503). Although I just noticed that this is already a known issue.