System Administration
cancel
Showing results for 
Search instead for 
Did you mean: 

clearing memory error count in cstm

GTarone
Occasional Contributor

clearing memory error count in cstm

We have an old I40 running HP-UX 10.20 & when we run cstm> il commnad we have a dimm with a large memory count error (+1000). We have replaced the dimm's but since the NOVA class doesn't support a page deallocation table we can't seem to reset the memory error count. We have re-mapped the system, reset the system activity log, cleared the tool status, cleared device status, update tool info even ran the exercise on memory & re-run the info tool but the count remains the same. Any help on how to clear the memory error log or if someone knows were it is I'll remove/clear it.
Thanks Gennaro T.
3 REPLIES
Matti_Kurkela
Honored Contributor

Re: clearing memory error count in cstm

DISCLAIMER: the answer below is based on a mixture of half-forgotten memories and educated guesses.

I don't know about I40 specifically, but at least on later models, the Page Deallocation Table is a hardware/firmware level feature and can be reset from the PDC boot prompt, not from the STM diagnostic utility.

Based on the NOVA class manuals available on www.openpa.net and www.hpmuseum.net, these old machines would seem to have a very limited selection of commands at the PDC prompt. If the PDT clearing command does not exist there, I'm not sure where it might be... perhaps in the Offline Diagnostic Environment (ODE)?

Before September 2005, the ODE was included in every HP-UX CD kit. It was accessible by booting the system from the Support Plus CD. Since then, it was made into a separate CD image that is downloadable from software.hp.com. (But for a vintage server like this, perhaps an old version would actually have a better chance of working correctly?)

For information about ODE:

http://docs.hp.com/en/diag.html#3 Offline Diagnostics
http://docs.hp.com/en/5992-6605/index.html
http://www13.itrc.hp.com/service/diag/aboutDiagPasswords.do (valid HP-UX support agreement required)

The MEM utility of the ODE sounds promising, but it is one of the password-protected ones... and it seems to be not publicly documented, either. Perhaps its Help menu will be descriptive enough?

MK
MK
GTarone
Occasional Contributor

Re: clearing memory error count in cstm

Youâ re correct about the BCH's SERvice menu to clear the PDT in models starting after the NOVA class (D & K etc). But the NOVA class has a limited "BCH" & the CM prompt has nothing either. I tried those first along with looking at the boot messages with different size memory (like in the K-class you get a message talking about a different size memory being detected & the PDT being reset. Thatâ s why I'm stuck with the STM commands. As for the ODE I'm very familiar with all the tools (I'm actually an old & embarrassed HP CE) & although the MEMtest looks promising the ODE tools do not affect the OS's error counts in STM. I want to try a totally different size memory & although I know I don't get the boot message about a new memory size being discovered I was hoping it would reset the error count as I know it does in the K-class etc. Unfortunately this is a remote site & the SA's want an answer before they take the sever down again & my I-class in the lab doesn't have memory errors to see if this would be effective.
If someone know the name & location of STM's memory error log (according to STM this is the source of the errors we are seeing) I could possibly clear/replace it. But so far I have been unsuccessful in locating it.
cnb
Honored Contributor

Re: clearing memory error count in cstm

Not sure about 10.20 but, from the 11.11 man page:

man memlogd

/dev/dmem - Memory device file
/var/stm/logs/sys/memlogd_activity_log - Activity log of memlogd
/var/stm/logs/os/memlog - Memory log file
/var/stm/data/daemon_list - List of daemons installed
/var/stm/data/key_file - OS resource key file


Use STM's logtool utility to clear/view

# cstm
cstm> ru logtool

Memory Commands

ClearLog (CL) - Clear the memory log file
SetTimer (ST) - Change the polling interval for logging memory errors
ViewDealloc (VDA) - View errors pertaining to deallocated memory errors
ViewDetail (VD) - View the entire memory log file


hth,