DB2 - L

Seems like our Operations staff is always wanting to speed up the time it takes to shut an LPAR down and get it back up. But I've noticed that Db2 subsystems are taking longer to shut down, even with MODE(FORCE). Automation often is forced to MODIFY IRLM to get the subsystems down, which isn't ideal, since it leaves things in an inconsistent state.

We've automated console dumps for Db2 taking more than a couple minutes to come down, and sent them to IBM - in most cases, IBM tells us that Db2 is shutting down normally, but is taking a long time to do so. My understanding is that the time is typically spent in:

* Externalizing Db2 buffer data - with buffers in the GBs/10GBs, it's often a lot to write out.

* Closing datasets - this is the big one, we run a DSMAX of 170,000 on our Primary Prod subsystems, and on occasion we do get up there with actual open datasets.

Anyone else seen Db2 take longer to shut down, prior to routine maintenance or IPL?

Asking because...I'm wondering if it's worth opening an RFE to provide some mechanism for "pre-shutdown". I..e if it's late at night and you're getting ready to IPL, manually issue a command to the subsystem to start asynchronous tasks to do things like close datasets not actively in-use, or writing all BP/GBP changed pages to disk. Obviously not something you'd fire off during work because of the performance implications, but it might be useful to give Db2 a "head-start" on shutting down pre-IPL, so the time from STOP DB2 to SHUTDOWN COMPLETE is minimal, without actually stopping service.

Anyone else noticed an uptick in shutdown times for a Db2 subsystem?

------------------------------
Mark Wieczorkowski
------------------
Db2 Systems Programmer, SSA/DCS
Principal - Solipsistic, LLC
------------------------------

Hello.

My suggestion is to force a checkpoint before issuing the stop db2 command.
This can shorten the time of the shutdown.

Regards,
Shay.

------------------------------
ShayMillerMataf
------------------------------

Original Message:
Sent: Apr 19, 2022 02:34 PM
From: Mark Wieczorkowski
Subject: Any other shops noticing Db2 taking a long time to shut down??

Seems like our Operations staff is always wanting to speed up the time it takes to shut an LPAR down and get it back up. But I've noticed that Db2 subsystems are taking longer to shut down, even with MODE(FORCE). Automation often is forced to MODIFY IRLM to get the subsystems down, which isn't ideal, since it leaves things in an inconsistent state.

We've automated console dumps for Db2 taking more than a couple minutes to come down, and sent them to IBM - in most cases, IBM tells us that Db2 is shutting down normally, but is taking a long time to do so. My understanding is that the time is typically spent in:

* Externalizing Db2 buffer data - with buffers in the GBs/10GBs, it's often a lot to write out.

* Closing datasets - this is the big one, we run a DSMAX of 170,000 on our Primary Prod subsystems, and on occasion we do get up there with actual open datasets.

Anyone else seen Db2 take longer to shut down, prior to routine maintenance or IPL?

Asking because...I'm wondering if it's worth opening an RFE to provide some mechanism for "pre-shutdown". I..e if it's late at night and you're getting ready to IPL, manually issue a command to the subsystem to start asynchronous tasks to do things like close datasets not actively in-use, or writing all BP/GBP changed pages to disk. Obviously not something you'd fire off during work because of the performance implications, but it might be useful to give Db2 a "head-start" on shutting down pre-IPL, so the time from STOP DB2 to SHUTDOWN COMPLETE is minimal, without actually stopping service.

Anyone else noticed an uptick in shutdown times for a Db2 subsystem?

------------------------------
Mark Wieczorkowski
------------------
Db2 Systems Programmer, SSA/DCS
Principal - Solipsistic, LLC
------------------------------

Original Message:
Sent: 4/19/2022 2:35:00 PM
From: Mark Wieczorkowski
Subject: Any other shops noticing Db2 taking a long time to shut down??

Seems like our Operations staff is always wanting to speed up the time it takes to shut an LPAR down and get it back up. But I've noticed that Db2 subsystems are taking longer to shut down, even with MODE(FORCE). Automation often is forced to MODIFY IRLM to get the subsystems down, which isn't ideal, since it leaves things in an inconsistent state.

We've automated console dumps for Db2 taking more than a couple minutes to come down, and sent them to IBM - in most cases, IBM tells us that Db2 is shutting down normally, but is taking a long time to do so. My understanding is that the time is typically spent in:

* Externalizing Db2 buffer data - with buffers in the GBs/10GBs, it's often a lot to write out.

* Closing datasets - this is the big one, we run a DSMAX of 170,000 on our Primary Prod subsystems, and on occasion we do get up there with actual open datasets.

Anyone else seen Db2 take longer to shut down, prior to routine maintenance or IPL?

Asking because...I'm wondering if it's worth opening an RFE to provide some mechanism for "pre-shutdown". I..e if it's late at night and you're getting ready to IPL, manually issue a command to the subsystem to start asynchronous tasks to do things like close datasets not actively in-use, or writing all BP/GBP changed pages to disk. Obviously not something you'd fire off during work because of the performance implications, but it might be useful to give Db2 a "head-start" on shutting down pre-IPL, so the time from STOP DB2 to SHUTDOWN COMPLETE is minimal, without actually stopping service.

Anyone else noticed an uptick in shutdown times for a Db2 subsystem?

------------------------------
Mark Wieczorkowski
------------------
Db2 Systems Programmer, SSA/DCS
Principal - Solipsistic, LLC
------------------------------

Original Message:
Sent: 4/20/2022 6:34:00 AM
From: Aurora Emanuela Dellanno
Subject: RE: Any other shops noticing Db2 taking a long time to shut down??

Hey Mark,

yes we have similar issues, and are trying to address them by shutting distributed access by stopping DDF well in advance of trying to stop Db2 - it doesn't always work and when we get to IPL time we sometimes have to wait ages...

I don't see how the "pre-shutdown" could possibly work, but I'm sure we would be interested in at least asking IBM to investigate what could be done.

Thanks.

Aurora

NB stay safe and healthy, y'all

Original Message:
Sent: 4/19/2022 2:35:00 PM
From: Mark Wieczorkowski
Subject: Any other shops noticing Db2 taking a long time to shut down??

Seems like our Operations staff is always wanting to speed up the time it takes to shut an LPAR down and get it back up. But I've noticed that Db2 subsystems are taking longer to shut down, even with MODE(FORCE). Automation often is forced to MODIFY IRLM to get the subsystems down, which isn't ideal, since it leaves things in an inconsistent state.

We've automated console dumps for Db2 taking more than a couple minutes to come down, and sent them to IBM - in most cases, IBM tells us that Db2 is shutting down normally, but is taking a long time to do so. My understanding is that the time is typically spent in:

* Externalizing Db2 buffer data - with buffers in the GBs/10GBs, it's often a lot to write out.

* Closing datasets - this is the big one, we run a DSMAX of 170,000 on our Primary Prod subsystems, and on occasion we do get up there with actual open datasets.

Anyone else seen Db2 take longer to shut down, prior to routine maintenance or IPL?

Asking because...I'm wondering if it's worth opening an RFE to provide some mechanism for "pre-shutdown". I..e if it's late at night and you're getting ready to IPL, manually issue a command to the subsystem to start asynchronous tasks to do things like close datasets not actively in-use, or writing all BP/GBP changed pages to disk. Obviously not something you'd fire off during work because of the performance implications, but it might be useful to give Db2 a "head-start" on shutting down pre-IPL, so the time from STOP DB2 to SHUTDOWN COMPLETE is minimal, without actually stopping service.

Anyone else noticed an uptick in shutdown times for a Db2 subsystem?

------------------------------
Mark Wieczorkowski
------------------
Db2 Systems Programmer, SSA/DCS
Principal - Solipsistic, LLC
------------------------------

We also issue STOP DDF, typically with MODE(FORCE), when shutting down a Db2 subsystem. DDF comes down quickly, so that's not our issue. We also issue STOP FUNC SPEC(*.*) and STOP PROC(*.*), since IBM recommended that action.

This delay appears due to actions taken to shut the subsystem itself down in a consistent state...mostly externalizing buffers and closing datasets.

------------------------------
Mark Wieczorkowski
------------------
Db2 Systems Programmer, SSA/DCS
Principal - Solipsistic, LLC
------------------------------

Original Message:
Sent: Apr 20, 2022 02:06 PM
From: Melissa Case
Subject: Any other shops noticing Db2 taking a long time to shut down??

Aurora –

It's been awhile!  I'm glad to see you're doing well!

 

We had a similar situation & have -STO DDF MODE(SUSPEND) CANCEL(5)

 

It works well for us.  There's a wait step, then a force DDF stop step.....  If the gentle way doesn't work, we go for the sledge hammer!

 

Best of luck

Missy Case

 

Original Message:
Sent: 4/20/2022 6:34:00 AM
From: Aurora Emanuela Dellanno
Subject: RE: Any other shops noticing Db2 taking a long time to shut down??

Hey Mark,

yes we have similar issues, and are trying to address them by shutting distributed access by stopping DDF well in advance of trying to stop Db2 - it doesn't always work and when we get to IPL time we sometimes have to wait ages...

I don't see how the "pre-shutdown" could possibly work, but I'm sure we would be interested in at least asking IBM to investigate what could be done.

Thanks.

Aurora

NB stay safe and healthy, y'all

Original Message:
Sent: 4/19/2022 2:35:00 PM
From: Mark Wieczorkowski
Subject: Any other shops noticing Db2 taking a long time to shut down??

Seems like our Operations staff is always wanting to speed up the time it takes to shut an LPAR down and get it back up. But I've noticed that Db2 subsystems are taking longer to shut down, even with MODE(FORCE). Automation often is forced to MODIFY IRLM to get the subsystems down, which isn't ideal, since it leaves things in an inconsistent state.

We've automated console dumps for Db2 taking more than a couple minutes to come down, and sent them to IBM - in most cases, IBM tells us that Db2 is shutting down normally, but is taking a long time to do so. My understanding is that the time is typically spent in:

* Externalizing Db2 buffer data - with buffers in the GBs/10GBs, it's often a lot to write out.

* Closing datasets - this is the big one, we run a DSMAX of 170,000 on our Primary Prod subsystems, and on occasion we do get up there with actual open datasets.

Anyone else seen Db2 take longer to shut down, prior to routine maintenance or IPL?

Asking because...I'm wondering if it's worth opening an RFE to provide some mechanism for "pre-shutdown". I..e if it's late at night and you're getting ready to IPL, manually issue a command to the subsystem to start asynchronous tasks to do things like close datasets not actively in-use, or writing all BP/GBP changed pages to disk. Obviously not something you'd fire off during work because of the performance implications, but it might be useful to give Db2 a "head-start" on shutting down pre-IPL, so the time from STOP DB2 to SHUTDOWN COMPLETE is minimal, without actually stopping service.

Anyone else noticed an uptick in shutdown times for a Db2 subsystem?

------------------------------
Mark Wieczorkowski
------------------
Db2 Systems Programmer, SSA/DCS
Principal - Solipsistic, LLC
------------------------------

Original Message:
Sent: 4/20/2022 2:11:00 PM
From: Mark Wieczorkowski
Subject: RE: Any other shops noticing Db2 taking a long time to shut down??

We also issue STOP DDF, typically with MODE(FORCE), when shutting down a Db2 subsystem. DDF comes down quickly, so that's not our issue. We also issue STOP FUNC SPEC(*.*) and STOP PROC(*.*), since IBM recommended that action.

This delay appears due to actions taken to shut the subsystem itself down in a consistent state...mostly externalizing buffers and closing datasets.

------------------------------
Mark Wieczorkowski
------------------
Db2 Systems Programmer, SSA/DCS
Principal - Solipsistic, LLC
------------------------------

Original Message:
Sent: Apr 20, 2022 02:06 PM
From: Melissa Case
Subject: Any other shops noticing Db2 taking a long time to shut down??

Aurora –

It's been awhile!  I'm glad to see you're doing well!

 

We had a similar situation & have -STO DDF MODE(SUSPEND) CANCEL(5)

 

It works well for us.  There's a wait step, then a force DDF stop step.....  If the gentle way doesn't work, we go for the sledge hammer!

 

Best of luck

Missy Case

 

Original Message:
Sent: 4/20/2022 6:34:00 AM
From: Aurora Emanuela Dellanno
Subject: RE: Any other shops noticing Db2 taking a long time to shut down??

Hey Mark,

yes we have similar issues, and are trying to address them by shutting distributed access by stopping DDF well in advance of trying to stop Db2 - it doesn't always work and when we get to IPL time we sometimes have to wait ages...

I don't see how the "pre-shutdown" could possibly work, but I'm sure we would be interested in at least asking IBM to investigate what could be done.

Thanks.

Aurora

NB stay safe and healthy, y'all

Original Message:
Sent: 4/19/2022 2:35:00 PM
From: Mark Wieczorkowski
Subject: Any other shops noticing Db2 taking a long time to shut down??

Seems like our Operations staff is always wanting to speed up the time it takes to shut an LPAR down and get it back up. But I've noticed that Db2 subsystems are taking longer to shut down, even with MODE(FORCE). Automation often is forced to MODIFY IRLM to get the subsystems down, which isn't ideal, since it leaves things in an inconsistent state.

We've automated console dumps for Db2 taking more than a couple minutes to come down, and sent them to IBM - in most cases, IBM tells us that Db2 is shutting down normally, but is taking a long time to do so. My understanding is that the time is typically spent in:

* Externalizing Db2 buffer data - with buffers in the GBs/10GBs, it's often a lot to write out.

* Closing datasets - this is the big one, we run a DSMAX of 170,000 on our Primary Prod subsystems, and on occasion we do get up there with actual open datasets.

Anyone else seen Db2 take longer to shut down, prior to routine maintenance or IPL?

Asking because...I'm wondering if it's worth opening an RFE to provide some mechanism for "pre-shutdown". I..e if it's late at night and you're getting ready to IPL, manually issue a command to the subsystem to start asynchronous tasks to do things like close datasets not actively in-use, or writing all BP/GBP changed pages to disk. Obviously not something you'd fire off during work because of the performance implications, but it might be useful to give Db2 a "head-start" on shutting down pre-IPL, so the time from STOP DB2 to SHUTDOWN COMPLETE is minimal, without actually stopping service.

Anyone else noticed an uptick in shutdown times for a Db2 subsystem?

------------------------------
Mark Wieczorkowski
------------------
Db2 Systems Programmer, SSA/DCS
Principal - Solipsistic, LLC
------------------------------

Original Message:
Sent: 4/20/2022 2:39:00 PM
From: Isaac Yassin
Subject: RE: Any other shops noticing Db2 taking a long time to shut down??

Hi, 

If your vdwqt (& classy  for ds) are low enough and chkp freq also low (4-5 min) then it's dataset close which is zos work. 

Isaac Yassin
Sent from my Note X

Original Message:
Sent: 4/20/2022 2:11:00 PM
From: Mark Wieczorkowski
Subject: RE: Any other shops noticing Db2 taking a long time to shut down??

We also issue STOP DDF, typically with MODE(FORCE), when shutting down a Db2 subsystem. DDF comes down quickly, so that's not our issue. We also issue STOP FUNC SPEC(*.*) and STOP PROC(*.*), since IBM recommended that action.

This delay appears due to actions taken to shut the subsystem itself down in a consistent state...mostly externalizing buffers and closing datasets.

------------------------------
Mark Wieczorkowski
------------------
Db2 Systems Programmer, SSA/DCS
Principal - Solipsistic, LLC
------------------------------

Original Message:
Sent: Apr 20, 2022 02:06 PM
From: Melissa Case
Subject: Any other shops noticing Db2 taking a long time to shut down??

Aurora –

It's been awhile!  I'm glad to see you're doing well!

 

We had a similar situation & have -STO DDF MODE(SUSPEND) CANCEL(5)

 

It works well for us.  There's a wait step, then a force DDF stop step.....  If the gentle way doesn't work, we go for the sledge hammer!

 

Best of luck

Missy Case

 

Original Message:
Sent: 4/20/2022 6:34:00 AM
From: Aurora Emanuela Dellanno
Subject: RE: Any other shops noticing Db2 taking a long time to shut down??

Hey Mark,

yes we have similar issues, and are trying to address them by shutting distributed access by stopping DDF well in advance of trying to stop Db2 - it doesn't always work and when we get to IPL time we sometimes have to wait ages...

I don't see how the "pre-shutdown" could possibly work, but I'm sure we would be interested in at least asking IBM to investigate what could be done.

Thanks.

Aurora

NB stay safe and healthy, y'all

Original Message:
Sent: 4/19/2022 2:35:00 PM
From: Mark Wieczorkowski
Subject: Any other shops noticing Db2 taking a long time to shut down??

Seems like our Operations staff is always wanting to speed up the time it takes to shut an LPAR down and get it back up. But I've noticed that Db2 subsystems are taking longer to shut down, even with MODE(FORCE). Automation often is forced to MODIFY IRLM to get the subsystems down, which isn't ideal, since it leaves things in an inconsistent state.

We've automated console dumps for Db2 taking more than a couple minutes to come down, and sent them to IBM - in most cases, IBM tells us that Db2 is shutting down normally, but is taking a long time to do so. My understanding is that the time is typically spent in:

* Externalizing Db2 buffer data - with buffers in the GBs/10GBs, it's often a lot to write out.

* Closing datasets - this is the big one, we run a DSMAX of 170,000 on our Primary Prod subsystems, and on occasion we do get up there with actual open datasets.

Anyone else seen Db2 take longer to shut down, prior to routine maintenance or IPL?

Asking because...I'm wondering if it's worth opening an RFE to provide some mechanism for "pre-shutdown". I..e if it's late at night and you're getting ready to IPL, manually issue a command to the subsystem to start asynchronous tasks to do things like close datasets not actively in-use, or writing all BP/GBP changed pages to disk. Obviously not something you'd fire off during work because of the performance implications, but it might be useful to give Db2 a "head-start" on shutting down pre-IPL, so the time from STOP DB2 to SHUTDOWN COMPLETE is minimal, without actually stopping service.

Anyone else noticed an uptick in shutdown times for a Db2 subsystem?

------------------------------
Mark Wieczorkowski
------------------
Db2 Systems Programmer, SSA/DCS
Principal - Solipsistic, LLC
------------------------------