Securepath agent CPU usage

[Carlos Ufano Gonzalez]

Hello, I have Securepath in Solaris 9 with two HBAs. The process spagent is using too much CPU laterly, and in some cases it have 80% of CPU usage.
Someone knows about these?

Regards, Carlos.

[Marco Hogeveen_1 · Accepted solution]

Hi Carlos,

This will be fixed in ServicePack1 for SP3.0D. Please contact HP to get it when it's released.

Marco

[Jan-Erik Karlsson_2]

It is possible to get a prefix which we got

Regards

[Carlos Ufano Gonzalez]

Please, could you send me that prefix? at this moment I have the spagent process with 70% of CPU.

[BMW 7x24 Support]

I have similar problems after Upgrading to 3.0D on Solaris 8. At some point in time the spagents on both Nodes of a VCS Cluster decide at the same time to go wild. Taking all CPU and communication with spmgr does not work. Restarting the spagents resolves the issue.

Two questions:

- is the trigger for this behaviour known?
- any idea when 3.0DSP1 will be available?

Thanks
Martin

[SAKET_5]

Hi Carlos,

Its interesting to read your post as when we reported this problem to HP atleast 3 months ago here in Brisbane, it took atleast a few weeks before they actually acknowledged that the issue is indeed a bug that they were not aware of. As I have gone through this pain of actually sending various log files on multiple times as our case escalated through numerous hops at HP, I try and present soem info based on memory recall regarding the case.

So, first just try to fire up a number of telnet sessions to your host running Secure Path V3.0D on its Secure Path agents ports (I forget the ports ...4991 or 4992...?..telnet hostIP:4991)...you would soon realise that just this is enough to make the spagent go wild and start eating away your CPU. It appears that the buggy spagent just doesnt know how to handle certain stages of TCP states and goes into a funny state where it just consumes CPU(CLOSE_WAIT, etc. mainly)...

Anyway...as we were concerned with the issue in an University environment where it is very possible that so many types of attacks could occur which can involve initiating a connection to the spagent's ports ....HP supplied us an updates spagent binary.

We have applied this binary although I am not sure if it has still gone through the proper HP testing and certifications requirement.

The updated binary in our experience has fixed the issue.

Hope it helps.

[Jakes Louw]

Hi guys

I'm working for Sun now (shoot me later), and I just picked up this problem on Solaris 9. Any progress on the issue? Links to the right places?