Hello,
as reported in many other threads, some UTM 220 models with only 2GB RAM suffer from massive swapping, causing overall slowdowns, lost packets, broken postgreSQL-DB and in the end defective HDDs.
With the latest updates, we where happy to see, that Sophos has made some improvements to memory management, because the usage dropped by about 30% to 50% (!) - seemingly.
But there was absolutely no speed improvement, and looking at the Kernel-log we saw that it still reports many problems with memory management because of the lack of available RAM.
So what has really happened?
Thanks god we are also monitoring the UTM over SNMP, and therefore we know that Sophos has "tuned" the graphs to not show the permanent overload anymore. It's not just a smoothing of peak values, the whole graph and statistics are absolutely worthless, because even monitoring on-device with atop clearly shows that there is no real background for this drop in memory usage.
This is an absolutely ridiculous move, and instead of making anything better, you now get the impression, that you can't trust any value in the GUI anymore.
A typical example:
MEMSwap.png
We know that before 10:30 there had been a massive problem with RAM overload (looking at the logfiles and SNMP sensor statistics), leading to a postgreSQL crash and a node takeover with subsequent postgre-self-healing.
Looking at the attached graph you get an impression about how much the GUI is lying.
Just to let you know - don't trust these statistics!
Best regards,
SM
This thread was automatically locked due to age.
