Process Mapping Logo

Process Mapping - Forums

Sharing 19 years of knowledge and experience

 
Metastorm BPM forums
Sign up Latest Topics
 
 
 


Reply
  Author   Comment  
thomas

Senior Member
Registered:
Posts: 79
Reply with quote  #1 
We migrated Metastorm to version 7.6 at a client recently. There's an issue with To Do lists not refreshing - and the system event log repeatedly shows this error:

Database exception while trying to raise alerts.  Operation: ExecuteSQL.  Code: State:S1T00,Native:0,Origin:[Microsoft][ODBC SQL Server Driver] Description: Timeout expired

Is there a reason, other than possible slow network / SQL server, that can cause this?
(The engine and DB are on different servers on same network)

0
Nappy

Avatar / Picture

Guru
Registered:
Posts: 1,087
Reply with quote  #2 

SynAttackProtect

The problem arises on a busy SQL system and some requests will randomly seem to be dropped with a General Network Error. 

 

To fix use regedit: 

 HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Tcpip\Parameters 

 Add a new DWORD key SynAttackProtect with value of 0.  Restart server. 

This has been a problem we had (with weird stuff going on)


__________________
Post an example, and we will have a much better idea what the problem is. In about 90% of posts, the problem is one of communication. Examples bridge that gap.

PS that's for V6/7 for V9 the problem is V9 at the moment ;)
0
Jerome

Avatar / Picture

Guru
Registered:
Posts: 5,507
Reply with quote  #3 
possinly an index is fragmmented ir missing after the migration. Rebuild indexes after deleting deletion alerts, perhaps.
__________________
Post an example, and we will have a much better idea what the problem is. In about 90% of posts, the problem is one of communication. Examples bridge that gap.
0
thomas

Senior Member
Registered:
Posts: 79
Reply with quote  #4 
Great, will give those suggestions a go.
@Nappy - will that registry entry be created on the server running SQL?
@Jerome - you reckon just the Alert related tables' indexes to rebuild?



0
Jerome

Avatar / Picture

Guru
Registered:
Posts: 5,507
Reply with quote  #5 
start wil ealert, but check all if possible.

Sorry about m/posts, using my phone.

__________________
Post an example, and we will have a much better idea what the problem is. In about 90% of posts, the problem is one of communication. Examples bridge that gap.
0
Nappy

Avatar / Picture

Guru
Registered:
Posts: 1,087
Reply with quote  #6 
The entry will most likely not be there.
I had a problem couple of years back and this fixed it.

Basically MS added a (UNDOCUMENTED!!) security "feature" that would drop packets without ANY logging so no way to find out.
In our case it was traffic from the ework server to database server (both on same internal subnet) being dropped.
Just add the key (it should not be a problem unless you suffer from flooding attacks, very unlikely on an internal server)

But probably Jerome's lead is better ;)

__________________
Post an example, and we will have a much better idea what the problem is. In about 90% of posts, the problem is one of communication. Examples bridge that gap.

PS that's for V6/7 for V9 the problem is V9 at the moment ;)
0
thomas

Senior Member
Registered:
Posts: 79
Reply with quote  #7 
Sigh, MS and their hidden 'features' ...
I rebuilt all them Indexes and created the registry entries Nappy suggested - still time outs though :(
I still suspect something's up with the indexes, might have to recreate them completely. I did a test - a simple select top 30 from the eAlert table takes almost 20 seconds... doesn't sound healthy..

EDIT: Thinking about it, I guess it will take long to do a Top 20 if I include a Order By clause since it has to first fetch all 6 million lines...

0
thomas

Senior Member
Registered:
Posts: 79
Reply with quote  #8 
Ok, timeouts seem to have stopped after creating some more indexes on the rest of the columns. So between that and the registry entry this seems to be solved... now to silence the nagging users complaining about supposed "new" errors... I have the before-migration designer log on my side though, muhaha
0
thomas

Senior Member
Registered:
Posts: 79
Reply with quote  #9 
Back on this topic again, after sorting out some other stuff.

This client seem to not get any alerts on anybody's ToDo or Watch Lists at all. I created a basic process to test the functionality, using both static and dynamic roles - the eAlert table never gets updated.

Has anyone had this issue before? What can we do in such a case?

0
fil6s

New Member
Registered:
Posts: 2
Reply with quote  #10 
Hi,

I am getting exactly the same as you, did you manage to fix this and is so how?

Phil G
0
thomas

Senior Member
Registered:
Posts: 79
Reply with quote  #11 
The issue was with a certain process causing a permanent lock in the eAlertGeneratorLock. When an alert got stuck here, no any other alerts from the eAlert table would be processed, hence no items appearing on the ToDo and WatchLists. 
Eventually we just removed the process causing the locks, as it had some 3rd party integration component which wasn't supported anymore.

So start by checking if your eAlertGeneratorLock table has had something stuck in there for a while...
0
Jerome

Avatar / Picture

Guru
Registered:
Posts: 5,507
Reply with quote  #12 
We've had that a few times. I fail to see why it cannot be managed by the engine. It is clever enough to record the lock, but then gets stuck. Seems silly to me.

__________________
Post an example, and we will have a much better idea what the problem is. In about 90% of posts, the problem is one of communication. Examples bridge that gap.
0
thomas

Senior Member
Registered:
Posts: 79
Reply with quote  #13 
Definitely is a bit silly. We created a small monitoring process where the administrator could remove the locked event out, or it removes it after a set time. Else things just never gets processed if a stupid lock takes place...
0
Arnold

Member
Registered:
Posts: 48
Reply with quote  #14 
Hi,

I know it is quite old post but we faced this issue in production today and after investigating, I think we have found the root cause.

Checking the lines in the ealertrequest table corresponding to the 2 blocking lines in the ealertgenerator table (we have 2 engines), we realized that the columns ETODOLIST and EWATCHLIST were containing data like:
user1,user2,user3,user4user4user4,


Then checking the corresponding action on the process we realized that a bad role assignation was done on a variable. Something like:

%MyRole := %concatenate(%SelectSQL("select bpm_user from contact where tid='%PID' and bpm_role='SPM' and status=1",%CNXCLIENT),%USERNAME)

Basically is missing one coma.

We are able to reproduce it in a DEV environment.

I have opened a ticket to metastorm support, because I don't understand why this should block the whole alert generator.

I will let you know as soon as I have some feedback.

__________________
Arnold
0
Previous Topic | Next Topic
Print
Reply

Quick Navigation:


Create your own forum with Website Toolbox!