In 5.0.5 cluster the CPU can get to 100% load because of Syncthing although the manager is idle and has nothing to sync between the nodes.
Syncthing is configured to scan all folders all the time and not only the changes.
When configuring it to watch for changes and increasing the interval CPU load decrease to normal (1-2%)
Environment:
OS (CLI), HA cluster, cloud provider
------------------------------------
Steps to reproduce:
------------------
1.
2.
3.
Expected result:
---------------
Actual result:
-------------
Following the discussion on slack, we can change the syncthing config to be watching for changes and only do full rescans rarely (which syncthing's documentation suggests to do, in case the watcher isn't able to pick some changes up), by using a script like this:
This should be resolved both for 5.1 and as a patch for 5.0.5. I am adding to Sprint 4, but if we have spare resources earlier let’s look into it.
After discussing with , lets make patches for 5.0.5 and 4.6 as well
should we re-open this one or create clones for the other version patches?
my understanding the issue is on 4.6 and 5.0.5 if know any other versions that might have similar issues let’s create a separate issue as this one is closed.