08-16-2016 09:49 AM
Environment
HA Nodes = 2
Product = SFHA
Version 6.1.1
ERROR / WARNING
June 12 18:30:00 NODE4 kernel: LLT INFO V-14-1-10035 timer not called for 3235 ticks
June 12 18:30:00 NODE4 kernel: LLT INFO V-14-1-10205 link 0 (eth3) node 0 in trouble
Understanding
As per my understanding either HA node is heavily loaded or Network has delays. Furthermore HostMonitor log under /var/VRTSvcs/log shows high memory alerts like below
2016/xx/xx 18:30:01 VCS INFO V-16-10061-14064 HostMonitor:VCShm:monitor:Updating System attribute with Mem usage = 99%.
It seems evident that due to lack of hardware resources the above mentioned errors/warning occuring.
Contradiction
In parallel we already setup a crontab job which is redirecting SAR and free -m command output in a text file.on every couple of minutes, which doesnot show any hike in CPU & MEMORY. Means SAR result shows almost 95% idle and free -m command shows 80% free out of 10GB memory.
Query
- Could not understand why errors/warning happening if system resources are almost idle.
- Why HostMonitor log showing wrong result. As HostMonitor shows high memory and free -m command shows enough available memory.
08-16-2016 11:17 PM
You never responded over here:
From where VCS Host Monitor logs picks result ?
Have you tried the following as per the 1st URL in my post?
Linux
Compare the usage reported by HostMonitor with the info found in the /proc/stat file
08-17-2016 06:58 AM - edited 08-17-2016 07:14 AM
Thanks for highlighting my earlier post and suggested article on that post.
My question is remain same. Why free -m command output is not similar with HostMonitor log. HostMonitor keep showing memory crunch. However free -m command output shows plenty of memory under+/- cache/buffer > free column
OR, How can we extract free memory from /proc/stat ?
08-18-2016 04:01 AM
The answer is the same.
Compare the usage reported by HostMonitor with the info found in the /proc/stat file
If you feel that this is not the correct way to determine system resources, get in touch with Support and Product Management.
08-18-2016 04:19 AM
Yes. I am assuming that I need to log a case. Last days research elaborated me that /proc/stat is only for CPU.
access.redhat.com/discussions/2533881#comment-1087131