We have this problem with our cluster: every time we start importing a large amount of data, the WMI process on our Windows machines tries to use as much processor power as possible. For example, now the java cassandra process and the WMI process take 50% of the processor resources. As a result, the nodes die at random, and cluster performance seems to be deteriorating. Is there a way to prevent the use of these two processes for resources?
We use 64-bit Windows 2012R2 servers running on Intel Xeons 2.4 GHz and 8 GB processors. The cluster consists of three identical machines.
One short word of explanation - sometimes it is like this, for hours, sometimes after a couple of minutes everything calms down. Also, sometimes after a dormant period, the nuts go completely between the two nodes, and the processor is occupied by 99% of WMI and cassandra.
source
share