1
SmarterMail Enterprise 100.0.9210.28175 (Mar 20, 2025) Build 9210 memory leak
Problem reported by Scott Johnson - 4/3/2025 at 1:51 PM
Submitted
ALL versions created this year (2025) run on Linux (have not tested Windows in 2025) exhibit memory leaks.  We have had ticket open for months.  Current version is the worst of the group.  I have seen others post on this but it is not getting the appropriate attention.  When paid for software is having to be rebooted every 6 hours on a 64GB 20 core system there is an issue that should not be publish as a general release to the public.

What does it take to get this fixed?

18 Replies

Reply to Thread
0
Brian Bjerring-Jensen Replied
Do you have any addons running with smartermail?
2
Derek Curtis Replied
Employee Post
You're having this issue, but not every Linux user is having it. We're running Linux, and we're constantly updating with newer, non-public builds and whatnot, but our memory is fine. We have multiple other servers running Linux versions and they run fine.

So, it's something unique to your particular environment, or an account (or multiple accounts) or a domain, or a migration, or indexing, or something. That's why it's important for you to work with us and go through various troubleshooting steps. If we can replicate the issue(s) -- which we've been unable to do up to this point -- that goes a long way in helping us figure out the "why" of what's happening. 
Derek Curtis COO SmarterTools Inc. www.smartertools.com
0
Scott Johnson Replied
Your memory is not fine.  Server ran for years on windows.  There is an issue.  This is exactly the problem.  Watched many of these posts and the standard is to blame the user.

As to the question of addons.  Nothing from the marketplace, but it was a replacement for an Exchange sever so EWS/EAS/MAPI.  It is here where the memory issue is.  The more of these users the faster the memory goes. No POP or WebDav either.
0
Jay Dubb Replied
Scott, just a question, are you sure it's a memory leak and not just a ton of consumption from the EWS/EAS/MAPI protocols?  On a server with a little over 2,000 users, most using EAS and MAPI, we regularly saw memory consumption in the 80-90 GB range.  Things got a little better with the newer versions running later .NET framework versions, but the lowest on a regular weekday-- high 60s to mid 70s GB and there were no other apps or plug-ins running.  Just straight Smartermail and nothing else.
 
0
Scott Johnson Replied
Server crashes no matter the memory.  It has only 17 total users. Yes it is all related to  EWS/EAS/MAPI.  128 or 256G still consumes all memory and crashes. Consumes over 30gb of ram in 6 hours.  So far with support pulling all of the configs no one has found any of the alleged user configuration issues.  Running the DotNET garbage collection intervals as aggressive as possible makes no difference.  GC cannot recover leaked memory.  
4
Gabriele Maoret - SERSIS Replied
if your server crashes with only 17 users, it's really something weird in your environment...

We have thousands of users on 6 separate servers (hundreds of them using MAPI/EWS and EAS) and no problems at all with 32GB max on each server...
Gabriele Maoret - Head of SysAdmins and CISO at SERSIS Currently manages 6 SmarterMail installations (1 in the cloud for SERSIS which provides services to a few hundred third-party email domains + 5 on-premise for customers who prefer to have their mail server in-house)
0
Scott Johnson Replied
On Linux?
1
Gabriele Maoret - SERSIS Replied
actually no... 5 are Windows and only one is Linux.

Are your problems only with Linux?
Gabriele Maoret - Head of SysAdmins and CISO at SERSIS Currently manages 6 SmarterMail installations (1 in the cloud for SERSIS which provides services to a few hundred third-party email domains + 5 on-premise for customers who prefer to have their mail server in-house)
0
Oliver Replied
@Gabriele, may I ask which hard disks you are using? HHD, SSD or NVMe and which RAID level?
1
Scott Johnson Replied
My original post says Linux.  Devil is in the details.

To be very clear.  New Debian install, new Smartermail install, migrated from an existing windows server per instructions of Smartertools.  Ran without issue.  As soon as MAPI/EWS and EAS were used with newly added domains the issue started.  Stop using those and the issue stops.

So the thing about Linux, not universal like windows.  This server is:

Description:    Debian GNU/Linux 12 (bookworm)
Linux version 6.1.0-32-amd64 (gcc-12 (Debian 12.2.0-14) 12.2.0
glibc Version: 2.36-9+deb12u10

This may or may not be the same as others.  It is also of interest that Smartertools has chosen NOT to install standard dotNet packages as instructed by Microsoft.  So every system is different.  Having now written software for over 40 years I can assure you that this is a memory leak.  It is at the core of MAPI/EWS and EAS.  It is reproducible.  It is predictable and after Smartertools looking at the server they could not find a configuration issue/error.  

This does not mean that others will experience the issue or that Windows servers exhibit the issue.  Windows and Linux memory management sub-systems are entirely different  All of the kernel work we did in Linux during the mid-2000s taught us that.  Much of the code we brough from other platforms to Linux had issues.  

When users purchase software, pay for maintenance and pay for support they should not be subjected to being the debug platform.  This type of issue would have caused a failing grade in my CS classes 43 years ago.  Free you get what you pay for.  It is great that that software is working on many platforms.  I hope that continues.

No user or environment issue should cause these.

Feb 01 22:41:43 mail kernel: .NET TP Worker invoked oom-killer: gfp_mask=0x140cca(GFP_HIGHUSER_MOVABLE|__GFP_COMP), order=0, oom_score_adj=0
Feb 01 22:41:43 mail kernel: oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=/,mems_allowed=0,global_oom,task_memcg=/system.slice/smartermail.service,task=MailService,pid=817,uid=0
Feb 01 22:41:43 mail kernel: Out of memory: Killed process 817 (MailService) total-vm:309299092kB, anon-rss:30568144kB, file-rss:0kB, shmem-rss:153620kB, UID:0 pgtables:64220kB oom_score_adj:0


0
Matt Petty Replied
Employee Post
Could try adding some swap to the server, I think this is something I tried on some ubuntu servers just adding like 4gb/8gb helped a lot, its something windows has by default but later ubuntu/debians dont have it.
https://askubuntu.com/questions/178712/how-to-increase-swap-space
Matt Petty Senior Software Developer SmarterTools Inc. www.smartertools.com
2
Gabriele Maoret - SERSIS Replied
@Oliver: all disks are NVMe Enterprise. We use VMs on RaidZ1 Storage in ProxMox Clusters
Gabriele Maoret - Head of SysAdmins and CISO at SERSIS Currently manages 6 SmarterMail installations (1 in the cloud for SERSIS which provides services to a few hundred third-party email domains + 5 on-premise for customers who prefer to have their mail server in-house)
0
Brian Bjerring-Jensen Replied
0
Oliver Replied
@Gabriele: THX
0
Gabriele Maoret - SERSIS Replied
@Scott Johnson : I'm monitoring the only Linux distribution I have.
This is a new test installation with only two users enabled for now (more will be enabled in a week or two)...
I'll give you some updates in a few days (10-15)...
Gabriele Maoret - Head of SysAdmins and CISO at SERSIS Currently manages 6 SmarterMail installations (1 in the cloud for SERSIS which provides services to a few hundred third-party email domains + 5 on-premise for customers who prefer to have their mail server in-house)
1
Jason Replied

We're experiencing the same memory leak issue and are now on build 9224. We've had a support ticket open for over 5 months with no resolution. We're also running SmarterMail Enterprise, and originally reported the problem when our server was running Windows. Since then, we’ve migrated to Linux, but the issue still persists. It's extremely frustrating, especially given the lack of meaningful updates from support. I'm glad to see we're not the only ones noticing this. Hopefully, increased visibility here will help move things forward, as this is now holding up other server migrations.

0
Mart Vernik Replied
We are running SM at Debian 12, ~200 users, Monthly Avarage Hardware Usage at System Administrator Reports page shows 2GB. We currently have ~10 MAPI/EWS/EAS users. Don’t know if it makse differents but we are running Caddy in front of Kestrel.
0
Scott Johnson Replied
@Mart Vernik that is very different from the basic Smartermail install which uses the built-in web server.  However, it is a difference I did not think of when migrating from our Windows systems which were using IIS. 

Reply to Thread

Enter the verification text