Re: Corrupt Environment

Question

Hi Pascal,
We ran into this problem frequently.
It looked like these partitions got corrupt when they were running
a task that did not return to the client (eg an infinite loop or a very
long
request) and the client was shut down harshly.
Then it even caused EConsole to either hang or blow up when
trying to open the Forte_Executor.
When running from
partition workshop and when the application is hanging, killing harshly
the Forte client is the worst thing to do. Developers should always
try 'Cancel Run' and/or closing the partition workshop, or anything
but killing the Forte client.
The only work-around we found is to try to guess what process
is corrupt (ps -ef might help on a Unix server) and kill it manually.
This worked in 50% cases. Otherwise we had to kill all Forte
processes and recycle the node like you describe it.
I had sent a message on this forum a while ago (may be a
year) about this problem, but had no answer then. May be
someone has a clue by now ?
Vincent Figari
On Fri, 5 Jun 1998 11:40:17 +0200 "Rottier, Pascal"
<[email protected]> writes:
Hi all,
This is a problem that keeps popping up now an again.
When you look at the contents of a node in the envi-
ronment, you will see a "Forte executor". If you open
this, it contains a number of partitions. Some of these
may be "none", meaning the processes are still alive,
but they are not in use right now. Others are busy,
and have a name like "<ProjectName>_cl0_part3".
Somehow, some of these partitions have become
corrupt. You can't open them, you can't shut them
down. If you startup your application (in development),
it may try to use these partitions and hang.
The only solution we have is shutdown the entire node,
kill all processes manually that somehow won't go down
and start everything up again.
Here are my questions.
1) What went wrong?
2) How do we prevent this?
3) Is there another way to solve it, if preventing it failed?
Pascal Rottier.
To unsubscribe, email '[email protected]' with
'unsubscribe forte-users' as the body of the message.
Searchable thread archive
<URL:http://pinehurst.sageit.com/listarchive/>
You don't need to buy Internet access to use free Internet e-mail.
Get completely free e-mail from Juno at http://www.juno.com
Or call Juno at (800) 654-JUNO [654-5866]
To unsubscribe, email '[email protected]' with
'unsubscribe forte-users' as the body of the message.
Searchable thread archive <URL:http://pinehurst.sageit.com/listarchive/>

Answer 1

[Pascal Rottier]
The "none" ftexecs aren't the problem. They stay none and they
stay alive
for days if you don't cancel them. They just sit there, waiting
to be used.
It's like it was posted in a previous message. If the ftexec on
the server
sides cancels a bit too awkward, it hang's and doesn't respond
anymore.
When econsole of escript tries to open this ftexec, it waits
forever.
So it's not a "none" ftexec, but a running ftexec that's
cancelled, that causes
the problem.
Pascal.
Attached is a program that we wrote for killing those "None" ftexecs
every 15
minutes (hopefully, before they become corrupt). This is completely
based on
technote 11435, but a member of our staff added a service object to do
the
dirty work every 15 minutes. Also, there is a constant defined
(kServerNode)
that you'll want to set to the name of your server node.
-Martin-
To unsubscribe, email '[email protected]' with
'unsubscribe forte-users' as the body of the message.
Searchable thread archive <URL:http://pinehurst.sageit.com/listarchive/>