Hey community,
This week, I encountered an issue with AIX agent backups. I have several AIX servers running versions 7.2 and 7.3. The backup jobs start, but there's a significant delay before they actually begin—when they start at all. When they do, the throughput is extremely low, and some jobs run for over 20 hours. The most severe issues seem to occur on AIX 7.3.
To troubleshoot, I tried:
-
Uninstalling and reinstalling the Veeam Agent
-
Creating a new protection group to ensure fresh package deployment and a new XML file
However, the issue persists.
When the job starts, it returns the path to the log file, and upon reviewing the logs, I found that:
- Path
/var/log/veeam/Backup/JOB-Backup/Session_20250402_160020_\{311f3273-44bf-476e-93ab-7e8f9a943b07\}/Agent.Source.log
- Log
[02.04.2025 16:02:46.438] <53084711> fb | Estimation:
[02.04.2025 16:02:46.438] <53084711> fb | Backup stats:
[02.04.2025 16:02:46.438] <53084711> fb | Total counts: files [607165], dirs [60021], links[9595], hard_links[3583]
[02.04.2025 16:02:46.438] <53084711> fb | Sizes: total files size
[271919853250], max file size [40719097856], aver. file size [447851.660175], total dirs size [676930], max dir size [21859], aver. dir size [11.278219]
[02.04.2025 16:02:46.438] <53084711> fb | Files distribution: <1Kb [325475], 1-20Kb [230080], 20-50Kb [22112], 50-200Kb [16239], 200Kb-1Mb [8243], 1-100Mb [4717], 100Mb-1Gb [269], >1Gb [30] [02.04.2025 16:02:46.438] <53084711> fb | Dirs distribution: <10 [50076], 10-100 [9217], 100-1000 [701], 1000-10000 [25], >10000 [2]
[02.04.2025 16:03:16.777] <39977555> cli | WARN|Waiting until the size of mount cache will be less than threshold
[02.04.2025 16:04:16.777] <39977555> cli | WARN|Waiting until the size of mount cache will be less than threshold
I have found where if AIX agent is running 4.5 or 4.6 on AIX 7.2 and later the performance issue can be encountered. These agent versions will use ipV6 sockets. Currently the agent does not have a work around to only use ipV4 sockets, but it can be addressed at the AIX OS layer by disabling mtu_bypass/largesend.
First you can see at aix if the mtu_bypass are “on”, using:
root@sappiqas (/)# lsattr -El en0 -a mtu_bypass
mtu_bypass on Enable/Disable largesend for virtual Ethernet True
Then, disable it using the following command:
root@sappiqas (/)# chdev -l en0 -a mtu_bypass=off
en0 changed
**Before you turn it off, ensure that no service is using mtu_bypass.
After that, the agent job runs just fine.
