tusc: process 23230 ("process name"):Cannot currently attach to deactivated processes.

[Nagaraj komarsamy]

Hi all,
When i attach a process to tusc it shows the following error.
tusc: process 23230 ("process name"):Cannot currently attach to deactivated processes.
What are all the reason that the tusc to report such an error?
What is mean by deactivated process?
Some times the following error also occurs.
tusc: no process to attach to
Whats mean by it?
Please help to understand it better.

thanks
Naga

[Dennis Handly]

>What are all the reason that the tusc to report such an error?

What PID did you give tusc? A zombie?

>tusc: no process to attach to
>Whats mean by it?

The obvious, that PID doesn't exist.

[Don Morris_1]

A deactivated process is one that the system has removed from scheduling (i.e. made not runnable / not active) in order to page the entire process out to the swap device [there's metadata affiliated with the process that can be swapped as well when this is done, as opposed to just the working set pages]). I would expect your system is or was under memory pressure when this occurs as a result. (Processes will be reactivated when the pressure is resolved [if ever], though this is not instantaneous of course).

I couldn't say for sure on all the cases for "no process to attach to" although the obvious cases that come to mind are if you mistype the target PID or the process has already exited. In both cases, there wouldn't be a valid target for tusc to work with.

[Nagaraj komarsamy]

Hi,

Thanks for your help.
It seems that the process exists with the same process ID throught for many days. It has not get killed or not restarted.
But during a particular period of one hour duration the tusc has given the following error. And the CPU usage of the process got reduced to almost 0%. After that one hour period the CPU usage increased to 85%.
The CPU usage is recored as shown below.
> 16:10 1269M 56.72
> 16:15 1269M 69.79
> 16:20 1269M 65.85
> 16:25 1269M 0.02
The problamatic 1 hour when tusc given the following error.

> 17:20 1269M 83.77
> 17:25 1269M 0.07
> 17:30 1269M 1.49
> 17:35 1269M 0.11



process 23230 ("process name"):Cannot currently attach to deactivated processes.

As you said if the process has been deactivated temporarily, how can we confirm this? Is there any system log/trace files having this information?


Regards,
Naga

[Dennis Handly]

>if the process has been deactivated temporarily, how can we confirm this? Is there any system log/trace files having this information?

The confirmation IS the tusc error.
Why should this be logged? This is a normal activity when the system is under severe memory pressure.
How much memory do you have? What does "swapinfo -tam" show?

[Don Morris_1]

Agreed... I'd be considering what else runs in that hour (timed backup which might eat a bunch of memory for I/O? Some type of business logic doing data analysis?) and whether that amount of pressure is acceptable for that time period (or if you need to tune the pressure causing workload, the system in general or just add more RAM).

As far as distinguishing deactivated processes (besides just tusc telling you this, which I agree is clear enough), such processes will lack the PS_INCORE flag in the pst_flag field of the pst_status structure filled in via pstat_getproc(). I don't see a mention of this in the top manpage -- but ps or other pstat-based tools may report it. I would certainly expect Glance to do this (and running Glance around the memory pressure time to see where the physical memory pressure is coming from is likely also a good idea).

[Nagaraj komarsamy]

Hi,

Thanks for your help.
Is there any way of system logs, commands or utilities to find out a particular process has been deactivated during a particular period of time? or is there no way to find out this? We asking this because we can't easily reproduce this scenario to convince people.


Regards,
Naga

[Dennis Handly]

>Is there any way of system logs, commands or utilities to find out a particular process has been deactivated during a particular period of time? or is there no way to find out this?

You can use sar(1m) and sa1(1m) to monitor your system for the next time.
In particular "sar -w".

http://docs.hp.com/en/B2355-60130/sar.1M.html
http://docs.hp.com/en/B2355-60130/sa1.1M.html
http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=1262801
http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=1246070

Unfortunately it won't mention which process.

>we can't easily reproduce this scenario to convince people.

Convince them to buy more hardware?