EAServer 5.5 problem

Hi,

I'm trying to upgrade our TEST EAServer to 5.5 (Win 2003 server,
Oracle DB).
We're able to simulate user transactions by sending HTTP requests to
the ISS server pointed towards the EAServer.

If the CPU load of the "jagsrv.exe" is less than 50% in "windows task
manager", the server runs for days.
But if there is more than 50% CPU of the "jagsrv.exe" the EAServer
freezes after a few hours :
-	the logs are not updated (no sign of a problem at all in the logs).
-	the CPU usage of the "jagsrv.exe" drops to 0 .
-	it is not possible to access the EAServer by the "EAServere
manager".

Each time the EAServer froze the values at the "Windows task manager"
of the "mem usage" and "vm size" were different.

My questions are:
-	What is the recommended limit of CPU usage for EAServer 5.5?
-	What other parameters should be monitored to find the problem?


Thank  you,
Michael

0
Michael
5/31/2008 6:58:56 AM
📁 sybase.easerver.general
📃 7903 articles.
⭐ 0 followers.

💬 8 Replies
👁️‍🗨️ 834 Views

[email protected] wrote...
> Hi,
> 
> I'm trying to upgrade our TEST EAServer to 5.5 (Win 2003 server,
> Oracle DB).
> We're able to simulate user transactions by sending HTTP requests to
> the ISS server pointed towards the EAServer.
> 
> If the CPU load of the "jagsrv.exe" is less than 50% in "windows task
> manager", the server runs for days.
> But if there is more than 50% CPU of the "jagsrv.exe" the EAServer
> freezes after a few hours :
> -	the logs are not updated (no sign of a problem at all in the logs).
> -	the CPU usage of the "jagsrv.exe" drops to 0 .
> -	it is not possible to access the EAServer by the "EAServere
> manager".
> 
> Each time the EAServer froze the values at the "Windows task manager"
> of the "mem usage" and "vm size" were different.
> 
> My questions are:
> -	What is the recommended limit of CPU usage for EAServer 5.5?
> -	What other parameters should be monitored to find the problem?

How many concurrent users are you allowing?  Perhaps it is too high?  Or, maybe your 
database is limiting the number of connections.  I have seen something like this happen 
before.  Since you didn't mention them, it doesn't sound like you are getting memory alarms 
in the log.  I usually see these from my PowerBuilder components when I push things too far 
(in terms of using memory in very inefficient ways).
-- 
Jim Egan [TeamSybase]
0
Jim
6/2/2008 4:28:43 AM
On Jun 2, 7:28=A0am, Jim Egan   wrote:
> [email protected] wrote...
> > Hi,
>
> > I'm trying to upgrade our TEST EAServer to 5.5 (Win 2003 server,
> > Oracle DB).
> > We're able to simulate user transactions by sending HTTP requests to
> > the ISS server pointed towards the EAServer.
>
> > If the CPU load of the "jagsrv.exe" is less than 50% in "windows task
> > manager", the server runs for days.
> > But if there is more than 50% CPU of the "jagsrv.exe" the EAServer
> > freezes after a few hours :
> > - =A0the logs are not updated (no sign of a problem at all in the logs).=

> > - =A0the CPU usage of the "jagsrv.exe" drops to 0 .
> > - =A0it is not possible to access the EAServer by the "EAServere
> > manager".
>
> > Each time the EAServer froze the values at the "Windows task manager"
> > of the "mem usage" and "vm size" were different.
>
> > My questions are:
> > - =A0What is the recommended limit of CPU usage for EAServer 5.5?
> > - =A0What other parameters should be monitored to find the problem?
>
> How many concurrent users are you allowing? =A0Perhaps it is too high? =A0=
Or, maybe your
> database is limiting the number of connections. =A0I have seen something l=
ike this happen
> before. =A0Since you didn't mention them, it doesn't sound like you are ge=
tting memory alarms
> in the log. =A0I usually see these from my PowerBuilder components when I =
push things too far
> (in terms of using memory in very inefficient ways).
> --
> Jim Egan [TeamSybase]- Hide quoted text -
>
> - Show quoted text -

Jim,

You're right; there are no memory alarms in the log.
The number of connections to the database is low (around 2 connections
per 2-3 seconds).
I have two "users", which are two Unix scripts.
The scripts are doing this:
- send an HTTP request.
- wait for the answer.
- sleep for one second.
- do it all over again.

How can I tell that =93I=92m pushing too far=94?
What are the areas that need to be checked in such cases and what
tools do you suggest using?

Thank you,
Michael
0
Michael
6/2/2008 5:00:57 AM
[email protected] wrote...
> On Jun 2, 7:28=A0am, Jim Egan   wrote:
> > [email protected] wrote...
> > > Hi,
> >
> > > I'm trying to upgrade our TEST EAServer to 5.5 (Win 2003 server,
> > > Oracle DB).
> > > We're able to simulate user transactions by sending HTTP requests to
> > > the ISS server pointed towards the EAServer.
> >
> > > If the CPU load of the "jagsrv.exe" is less than 50% in "windows task
> > > manager", the server runs for days.
> > > But if there is more than 50% CPU of the "jagsrv.exe" the EAServer
> > > freezes after a few hours :
> > > - =A0the logs are not updated (no sign of a problem at all in the logs).=
> 
> > > - =A0the CPU usage of the "jagsrv.exe" drops to 0 .
> > > - =A0it is not possible to access the EAServer by the "EAServere
> > > manager".
> >
> > > Each time the EAServer froze the values at the "Windows task manager"
> > > of the "mem usage" and "vm size" were different.
> >
> > > My questions are:
> > > - =A0What is the recommended limit of CPU usage for EAServer 5.5?
> > > - =A0What other parameters should be monitored to find the problem?
> >
> > How many concurrent users are you allowing? =A0Perhaps it is too high? =A0=
> Or, maybe your
> > database is limiting the number of connections. =A0I have seen something l=
> ike this happen
> > before. =A0Since you didn't mention them, it doesn't sound like you are ge=
> tting memory alarms
> > in the log. =A0I usually see these from my PowerBuilder components when I =
> push things too far
> > (in terms of using memory in very inefficient ways).
> > --
> > Jim Egan [TeamSybase]- Hide quoted text -
> >
> > - Show quoted text -
> 
> Jim,
> 
> You're right; there are no memory alarms in the log.
> The number of connections to the database is low (around 2 connections
> per 2-3 seconds).
> I have two "users", which are two Unix scripts.
> The scripts are doing this:
> - send an HTTP request.
> - wait for the answer.
> - sleep for one second.
> - do it all over again.
> 
> How can I tell that =93I=92m pushing too far=94?
> What are the areas that need to be checked in such cases and what
> tools do you suggest using?

I would first use the runtime monitory within Jaguar Manager to verify that your components 
are behaving like you expect.  Look for the number of active components to keep growing.  
Since you said you have two connections to the database it sounds like your components are 
committing OK and aren't holding connections.  That's a good start.

With your two "users" is this fire and forget?  Or is it serial (a new request cannot be 
made until the prior request is complete)?

Also check overall memory utilization within Windows.  Is it creeping up consistently over 
time?  If so then you may have a memory leak.  But without a memory alarm I'm not sure 
what's going on.
-- 
Jim Egan [TeamSybase]
0
Jim
6/4/2008 6:52:39 AM
On Jun 4, 9:52=A0am, Jim Egan   wrote:
> voic[email protected] wrote...
> > On Jun 2, 7:28=3DA0am, Jim Egan =A0 wrote:=

> > > [email protected] wrote...
> > > > Hi,
>
> > > > I'm trying to upgrade our TEST EAServer to 5.5 (Win 2003 server,
> > > > Oracle DB).
> > > > We're able to simulate user transactions by sending HTTP requests to=

> > > > the ISS server pointed towards the EAServer.
>
> > > > If the CPU load of the "jagsrv.exe" is less than 50% in "windows tas=
k
> > > > manager", the server runs for days.
> > > > But if there is more than 50% CPU of the "jagsrv.exe" the EAServer
> > > > freezes after a few hours :
> > > > - =3DA0the logs are not updated (no sign of a problem at all in the =
logs).=3D
>
> > > > - =3DA0the CPU usage of the "jagsrv.exe" drops to 0 .
> > > > - =3DA0it is not possible to access the EAServer by the "EAServere
> > > > manager".
>
> > > > Each time the EAServer froze the values at the "Windows task manager=
"
> > > > of the "mem usage" and "vm size" were different.
>
> > > > My questions are:
> > > > - =3DA0What is the recommended limit of CPU usage for EAServer 5.5?
> > > > - =3DA0What other parameters should be monitored to find the problem=
?
>
> > > How many concurrent users are you allowing? =3DA0Perhaps it is too hig=
h? =3DA0=3D
> > Or, maybe your
> > > database is limiting the number of connections. =3DA0I have seen somet=
hing l=3D
> > ike this happen
> > > before. =3DA0Since you didn't mention them, it doesn't sound like you =
are ge=3D
> > tting memory alarms
> > > in the log. =3DA0I usually see these from my PowerBuilder components w=
hen I =3D
> > push things too far
> > > (in terms of using memory in very inefficient ways).
> > > --
> > > Jim Egan [TeamSybase]- Hide quoted text -
>
> > > - Show quoted text -
>
> > Jim,
>
> > You're right; there are no memory alarms in the log.
> > The number of connections to the database is low (around 2 connections
> > per 2-3 seconds).
> > I have two "users", which are two Unix scripts.
> > The scripts are doing this:
> > - send an HTTP request.
> > - wait for the answer.
> > - sleep for one second.
> > - do it all over again.
>
> > How can I tell that =3D93I=3D92m pushing too far=3D94?
> > What are the areas that need to be checked in such cases and what
> > tools do you suggest using?
>
> I would first use the runtime monitory within Jaguar Manager to verify tha=
t your components
> are behaving like you expect. =A0Look for the number of active components =
to keep growing. =A0
> Since you said you have two connections to the database it sounds like you=
r components are
> committing OK and aren't holding connections. =A0That's a good start.
>
> With your two "users" is this fire and forget? =A0Or is it serial (a new r=
equest cannot be
> made until the prior request is complete)?
>
> Also check overall memory utilization within Windows. =A0Is it creeping up=
 consistently over
> time? =A0If so then you may have a memory leak. =A0But without a memory al=
arm I'm not sure
> what's going on.
> --
> Jim Egan [TeamSybase]- Hide quoted text -
>
> - Show quoted text -

What I=92m looking for is a name of a tool which can show the internal
Jaguar memory partitioning; what is given to PB, Java, =85 vs=92 what is
actually used.
Maybe this way we can find what needs to be tuned.
--
Michael
0
Michael
6/4/2008 11:39:53 AM
[email protected] wrote...
> What I=92m looking for is a name of a tool which can show the internal
> Jaguar memory partitioning; what is given to PB, Java, =85 vs=92 what is
> actually used.
> Maybe this way we can find what needs to be tuned.
> 

For that I can't help you.  Maybe Evan will pipe in here with something that will help.

For faster turnaround on answers you might want to open a case with Sybase Technical 
Support.
-- 
Jim Egan [TeamSybase]
0
Jim
6/5/2008 6:25:37 AM
For EAServer 6.X, using the console, on the property page for your 
Server, under Tasks, enable:

     Dump60SecondMemoryUsage
     SybHeap_dumpSummary
     PbHeap_dumpSummary

Then restart the server. Check the server log file for details.

sheet Jim Egan wrote:
> [email protected] wrote...
>> What I=92m looking for is a name of a tool which can show the internal
>> Jaguar memory partitioning; what is given to PB, Java, =85 vs=92 what is
>> actually used.
>> Maybe this way we can find what needs to be tuned.
>>
> 
> For that I can't help you.  Maybe Evan will pipe in here with something that will help.
> 
> For faster turnaround on answers you might want to open a case with Sybase Technical 
> Support.
0
Evan
6/6/2008 2:13:53 AM
On Jun 6, 4:13=A0am, Evan Ireland  wrote:
> For EAServer 6.X, using the console, on the property page for your
> Server, under Tasks, enable:
>
> =A0 =A0 =A0Dump60SecondMemoryUsage
> =A0 =A0 =A0SybHeap_dumpSummary
> =A0 =A0 =A0PbHeap_dumpSummary
>
> Then restart the server. Check the server log file for details.
>
>
>
> sheet Jim Egan wrote:
> > [email protected] wrote...
> >> What I=3D92m looking for is a name of a tool which can show the interna=
l
> >> Jaguar memory partitioning; what is given to PB, Java, =3D85 vs=3D92 wh=
at is
> >> actually used.
> >> Maybe this way we can find what needs to be tuned.
>
> > For that I can't help you. =A0Maybe Evan will pipe in here with somethin=
g that will help.
>
> > For faster turnaround on answers you might want to open a case with Syba=
se Technical
> > Support.- Hide quoted text -
>
> - Show quoted text -

Hi,

I'm using EAServer 5.5.
Are there any suggestions/tools to tackle the memory issue on 5.5?

By the way, would it be a wise step to move from 5.5 to version 6.****?
If so, what is the best 6.**** version and what patches to use?

Thank you,
Michael
0
Michael
6/8/2008 4:42:38 AM
See http://www.sybase.com/detail?id=1027319

Michael wrote:
> On Jun 6, 4:13 am, Evan Ireland  wrote:
>> For EAServer 6.X, using the console, on the property page for your
>> Server, under Tasks, enable:
>>
>>      Dump60SecondMemoryUsage
>>      SybHeap_dumpSummary
>>      PbHeap_dumpSummary
>>
>> Then restart the server. Check the server log file for details.
>>
>>
>>
>> sheet Jim Egan wrote:
>>> [email protected] wrote...
>>>> What I=92m looking for is a name of a tool which can show the internal
>>>> Jaguar memory partitioning; what is given to PB, Java, =85 vs=92 what is
>>>> actually used.
>>>> Maybe this way we can find what needs to be tuned.
>>> For that I can't help you.  Maybe Evan will pipe in here with something that will help.
>>> For faster turnaround on answers you might want to open a case with Sybase Technical
>>> Support.- Hide quoted text -
>> - Show quoted text -
> 
> Hi,
> 
> I'm using EAServer 5.5.
> Are there any suggestions/tools to tackle the memory issue on 5.5?
> 
> By the way, would it be a wise step to move from 5.5 to version 6.****?
> If so, what is the best 6.**** version and what patches to use?
> 
> Thank you,
> Michael
0
Evan
6/9/2008 8:24:08 PM