MQ

Hello MQ community users,

Does anybody knows whether  a start/stop of TSM backup interface could actually impact MQ Server by dropping down MQ command server and therefore making unavailable Queue Managers?

The above seems to happened in our case when in a context of a project rehearsal the AIX admin has stopped / restarted TSM Backup service (using its own interface / IP) to perform scheduled backups on AIX (Active/ Passive Power HA based) MQ host and at exactly the same timestamp -as recorded in MQ error logs- MQ command server ended  causing all three (3) QMs hosted on this AIX clustered MQ host to become unavailable.

Actually the stop/start on the interface used for TSM Backup, when the interface was started again the hostname of the LPAR was changed to the standalone hostname(which was different than the one that was active when MQ services were started)

As a result the MQSeries could not find the file /var/mqm/sockets/NBG_T24/qmgrloc/T24MQ1/subpool.lck 

I am thinking of raising a case to IBM in order to track the root cause considering the remarkable impact in production environment.

If anybody has experienced a symptom like the one described above, i would appreciate for a feedback or advise about how to prevent it in future.

P.S.

Up to my knowledge it is not feasible to "bind" MQ service to specific network interface IP on AIX system that has multiple network interfaces available like in our case.

Thanks in advance

Cheers,  Nick.

------------------------------
Nick Dakoronias
------------------------------

Hi Nick,

It is a bad idea to change a hostname while MQ is running.

I remember seeing something (may be a technote) where the use of an environment variable pointing to the old hostname allows you to reaccess the queue manager (IPCCs). The queue manager is not unavailable, what is unavailable after the hostname change is the IPCC environment you need to talk to the queue manager.

Hope it helps

Hello MQ community users,

Does anybody knows whether  a start/stop of TSM backup interface could actually impact MQ Server by dropping down MQ command server and therefore making unavailable Queue Managers?

The above seems to happened in our case when in a context of a project rehearsal the AIX admin has stopped / restarted TSM Backup service (using its own interface / IP) to perform scheduled backups on AIX (Active/ Passive Power HA based) MQ host and at exactly the same timestamp -as recorded in MQ error logs- MQ command server ended  causing all three (3) QMs hosted on this AIX clustered MQ host to become unavailable.

Actually the stop/start on the interface used for TSM Backup, when the interface was started again the hostname of the LPAR was changed to the standalone hostname(which was different than the one that was active when MQ services were started)

As a result the MQSeries could not find the file /var/mqm/sockets/NBG_T24/qmgrloc/T24MQ1/subpool.lck 

I am thinking of raising a case to IBM in order to track the root cause considering the remarkable impact in production environment.

If anybody has experienced a symptom like the one described above, i would appreciate for a feedback or advise about how to prevent it in future.

P.S.

Up to my knowledge it is not feasible to "bind" MQ service to specific network interface IP on AIX system that has multiple network interfaces available like in our case.

Thanks in advance

Cheers,  Nick.

------------------------------
Nick Dakoronias
------------------------------

Hi Francois,

Thanks for your reply.

Regarding your hint about the environment variable, i assume you are implying the export MQS_IPC_HOST=<old host name>

Cheers, Nick.

------------------------------
Nick Dakoronias

Original Message:
Sent: Thu May 30, 2024 08:42 AM
From: Francois Brandelik
Subject: TSM backup interface start/stop can affect MQ availability?

Hi Nick,

It is a bad idea to change a hostname while MQ is running.

I remember seeing something (may be a technote) where the use of an environment variable pointing to the old hostname allows you to reaccess the queue manager (IPCCs). The queue manager is not unavailable, what is unavailable after the hostname change is the IPCC environment you need to talk to the queue manager.

Hope it helps

------------------------------
Francois Brandelik

Original Message:
Sent: Wed May 29, 2024 09:31 AM
From: NICK DAKORONIAS
Subject: TSM backup interface start/stop can affect MQ availability?

Hello MQ community users,

Does anybody knows whether  a start/stop of TSM backup interface could actually impact MQ Server by dropping down MQ command server and therefore making unavailable Queue Managers?

The above seems to happened in our case when in a context of a project rehearsal the AIX admin has stopped / restarted TSM Backup service (using its own interface / IP) to perform scheduled backups on AIX (Active/ Passive Power HA based) MQ host and at exactly the same timestamp -as recorded in MQ error logs- MQ command server ended  causing all three (3) QMs hosted on this AIX clustered MQ host to become unavailable.

Actually the stop/start on the interface used for TSM Backup, when the interface was started again the hostname of the LPAR was changed to the standalone hostname(which was different than the one that was active when MQ services were started)

As a result the MQSeries could not find the file /var/mqm/sockets/NBG_T24/qmgrloc/T24MQ1/subpool.lck 

I am thinking of raising a case to IBM in order to track the root cause considering the remarkable impact in production environment.

If anybody has experienced a symptom like the one described above, i would appreciate for a feedback or advise about how to prevent it in future.

P.S.

Up to my knowledge it is not feasible to "bind" MQ service to specific network interface IP on AIX system that has multiple network interfaces available like in our case.

Thanks in advance

Cheers,  Nick.

------------------------------
Nick Dakoronias
------------------------------

Exactly! Have you tried it?

As for your PS:

Try setting the local address both on the listener and outgoing channels.

This will bind the communications to said local IP (interface)...

------------------------------
Francois Brandelik

Original Message:
Sent: Fri May 31, 2024 05:58 AM
From: NICK DAKORONIAS
Subject: TSM backup interface start/stop can affect MQ availability?

Hi Francois,

Thanks for your reply.

Regarding your hint about the environment variable, i assume you are implying the export MQS_IPC_HOST=<old host name>

Cheers, Nick.

------------------------------
Nick Dakoronias

Original Message:
Sent: Thu May 30, 2024 08:42 AM
From: Francois Brandelik
Subject: TSM backup interface start/stop can affect MQ availability?

Hi Nick,

It is a bad idea to change a hostname while MQ is running.

I remember seeing something (may be a technote) where the use of an environment variable pointing to the old hostname allows you to reaccess the queue manager (IPCCs). The queue manager is not unavailable, what is unavailable after the hostname change is the IPCC environment you need to talk to the queue manager.

Hope it helps

------------------------------
Francois Brandelik

Original Message:
Sent: Wed May 29, 2024 09:31 AM
From: NICK DAKORONIAS
Subject: TSM backup interface start/stop can affect MQ availability?

Hello MQ community users,

Does anybody knows whether  a start/stop of TSM backup interface could actually impact MQ Server by dropping down MQ command server and therefore making unavailable Queue Managers?

The above seems to happened in our case when in a context of a project rehearsal the AIX admin has stopped / restarted TSM Backup service (using its own interface / IP) to perform scheduled backups on AIX (Active/ Passive Power HA based) MQ host and at exactly the same timestamp -as recorded in MQ error logs- MQ command server ended  causing all three (3) QMs hosted on this AIX clustered MQ host to become unavailable.

Actually the stop/start on the interface used for TSM Backup, when the interface was started again the hostname of the LPAR was changed to the standalone hostname(which was different than the one that was active when MQ services were started)

As a result the MQSeries could not find the file /var/mqm/sockets/NBG_T24/qmgrloc/T24MQ1/subpool.lck 

I am thinking of raising a case to IBM in order to track the root cause considering the remarkable impact in production environment.

If anybody has experienced a symptom like the one described above, i would appreciate for a feedback or advise about how to prevent it in future.

P.S.

Up to my knowledge it is not feasible to "bind" MQ service to specific network interface IP on AIX system that has multiple network interfaces available like in our case.

Thanks in advance

Cheers,  Nick.

------------------------------
Nick Dakoronias
------------------------------