Error database question

[Mike Campbell]

Cloverleaf 5.7, patch 2 on aix.

Simple question for Monday.  Is there a command you can run against the error database to get a count of the messages?  When I run the IDE to return all messages it times out and doesn’t give any info.  I’ve selected individual Destinations and it does return information.  

I’d like to determine where the errors are before I re-initialize the database.  

Thanks.

Mike C.

Topic

[David Barr]

I usually do something like:

hcidbdump -e | awk ‘{ print $9, $10 }’ | sort | uniq -c

This gives you a count of messages by thread.

[Robert Kersemakers]

On HP-UX I normally do

Code:

hcidbdump -e |wc -l

to get an indication of the number of errors. Subtract 8 from the result because of header/footerlines.

Zuyderland Medisch Centrum; Heerlen/Sittard; The Netherlands

[Mike Campbell]

Thanks gentlemen.  Those worked as designed.

[Alex Puzikov]

I do even easier. I add /dev/null (on UNIX) as a file to write data into

Code:

% hcidbdump -e -U ap /dev/null

Processed 214 messages.

Done.
%

Count that you see is you actual count – no need to calculate anything  🙂

[Robert Kersemakers]

Alex: if there had been a ‘Like’ button, I would have clicked it. 😀

Zuyderland Medisch Centrum; Heerlen/Sittard; The Netherlands

[Mike Campbell]

All really great ideas.  Thanks all.  

Now, just one other question.  We had an issue on Friday that sent A LOT of messages to the error database.  I believe these were all ACK and can be deleted.  However, when I try to delete them using the IDE, it freezes the threads and causes disruption.  [Sit down..there are about 1,364,700 messages for the affected thread.]  Is there anyway to delete those without disrupting messages, or just bite the bullet and stop everything and reinitialize the database.

Again, thanks.

[Robert Kersemakers]

You can delete error messages with

Code:

hcidbdump -e -D

This will delete ALL errormessages, so use other parameters to narrow your selection.

Use

Code:

hcidbdump —

to have a look at all the available options for this.

Zuyderland Medisch Centrum; Heerlen/Sittard; The Netherlands

[Richard Hart]

Hi Mike.

We use external monitoring and I check the size of the error DB. If the size is greater that the ‘initialised file’ size, then we have an error.

A subsequent check is made on the lockmgr ‘pid’ file and an alert is raised ONLY if the db file is newer than the pid file.

i.e.

If a message is sent to the error database, an alert is raised. The work instruction for the alert fixes the issue and touches the pid file to remove the alert.

We have many sites and this save a dump on each one which could potentially contain a number of messages.

[Michael Hertel]

You may be able to get away with:

hcidbdump -e -U MIKE -D

This will use a different user name during your operation.

[Author]

Replies