Pings slowing down

Ludo@Work · Post by **Ludo@Work** » Mon Nov 15, 2010 9:48 am

Hi,

It's been quite a happy while that I've been using HM now, until I recently moved onto a 2008r2 box. I'm currently on HM 8.64.

Something strange is happening : ping tests to different addresses gradually get slower and slower and slower, this in a nice steady linear sort of fashion. Over the last 15 days or so, on average, the reply values have increased by some 7-10ms : e.g. 2ms became 12ms, 10ms became 20ms.

TCP/UPD port tests e.g. are not affected.

It doesn't seem to make much difference whether it's an address or a hostname.

At times that HM returns too slow responses, replies from CLI issued at the same time are normal.

I rebooted that 2008r2 box today, and all the HM-values are back to their original values (i.e. the ones I expect).

It smells like a memory leak. However, there is plenty of memory in the box, and when checking, more than a gig was free. I never really checked before how much memory HM is actually consuming, but the current 18-19MB does not seem outrageous.

Has anybody seen something similar ?
Any tips for further investigation ?

Regards,

Ludo

KS-Soft · Post by **KS-Soft** » Mon Nov 15, 2010 12:56 pm

We had similar problem related to our router.
Do you have installed some antivirus monitors, personal firewall, content monitoring software? Non stanard winsock components?

Could you check resource usage for each process? You may use standard Windows Task Manager to check Handles, GDI and USER objects. What is the total resource usage on the system? How many handles/threads/GDI objects used by hostmon.exe process?

Regards
Alex

Ludo@Work · Post by **Ludo@Work** » Wed Nov 17, 2010 5:58 am

Alex, I well be keeping an eye on it for the next couple of days.

I really think it is a memory leak : it started yesterday with 18420kb, and today it is 20140 already ...

Typically, there are 11 threads linked to HM (as reported by Proces explorer - sysinternals) with a few extra from time to time.

Ludo

KS-Soft · Post by **KS-Soft** » Wed Nov 17, 2010 10:40 am

What exactly memory are you checking? "Mem usage" counter displayed by Task Manager? It shows "working set" memory, which means nothing.
==============
Working Set is the current size, in bytes, of the Working Set of this process. The Working Set is the set of memory pages touched recently by the threads in the process. If free memory in the computer is above a threshold, pages are left in the Working Set of a process even if they are not in use. When free memory falls below a threshold, pages are trimmed from Working Sets. If they are needed they will then be soft-faulted back into the Working Set before leaving main memory.
==============

Could you check resource usage for each process? You may use standard Windows Task Manager to check Handles, GDI and USER objects. What is the total resource usage on the system? How many handles/threads/GDI objects used by hostmon.exe process?

Regards
Alex

Ludo@Work · Post by **Ludo@Work** » Wed Nov 24, 2010 4:12 am

Hi, just an update on the subject ...

It turned out to be no memory leaks at all

Disabling the NIC autotuning seems to have improved the situation a lot - if not cured.

But it may be too soon to see.

From an elevated powershell : netsh interface tcp set global autotuning=disabled.

Ludo

KS-Soft · Post by **KS-Soft** » Wed Nov 24, 2010 8:08 pm

Oh, autotuning...
Good to know.

There are several articles from Microsoft related to other problems caused by TCP/IP Window Auto-Tuning feature
http://support.microsoft.com/kb/934430
http://support.microsoft.com/kb/935400
http://support.microsoft.com/kb/947239
http://support.microsoft.com/kb/951291

Regards
Alex