Hi guys,
I have a strange situation with two ASR1002 (03.02. - 15.1(S)1 accting as BRAS.
Each of them have around 6k users.
Everithing works well for 2 years till last sunday.
From last sunday I had a situation twice where cpu goes above 90% (vtemplate procces), during that time all users which are connected are working well but new users can not get authenticated during this period, although there is IP connectivity to radius server.
It is very hard to debug anything becasue cpu is to high.
After removing authentication for ppp sessions (virtual-template) cpu goes down to the normal rate, but users are not authenticated.
Last sunday this condition was for 3-4 hours constantly, and after reboot the boxes everything works well till this morning.
This morning I had same behavior, but no reboot was required.
After some time and removing/adding authentication for ppp it get back to normal.
There is no logs, errors or anything which could point to the cause of the problem.
We had some drops on access links with default configuration.
Hold-queue in and out for access links are now set to 4096.
What we can see on radius is that number of request in that period goes from normal 10/s to 100/s, radius can handle it, but it seems like ASR have some problems during that time.
I am thinking to implement CAC feature, but I am afraid it is something more than scalability issue.
Number of users did not increase to much in last 2 years, maybe 1,5 k.
If anyone had similiar problems with ASR boxes in same environment please share your experience.
Any help/advice would be very usefull.
Br,
Grgo
