Is it mandatory to stop sender channel during PI server downtime

[Former Member]

Hi All,

I have some doubts regarding PI Server downtime activities. Currently I am working on 7.31 Dual Stack but 95 % interfaces are in ICO.

Ques :-  It is mandatory to stop all sender channel ?

Ans :-  Yes since Data will be lost

Doubt :- Where Data will be lost we have three cases .

1. Message are picking from sender side

2. Message are processing in PI

3.Message are receiving in Receiver.

In all the cases if we will not stop the channels but stop server it self , message will be failed in  sender and receiver system or in PI system

1. For all sender /receiver  system have these configuration either message will processed completely or roll back .

2. If Message will fail in PI we can reprocess that message

So now the question, is it still important to stop sender and receiver channel if yes then why ?

Thanks & Regards,

Rudra

[former_member184720 · Accepted solution]

Hi Rudra - It's not mandatory to stop the channels but in a productive environment it's always good to stop the channels.

I don't think it's not just about data loss instead i feel it's just another factor.

When you restart the server, it's always not guaranteed that all the PI components would be up and running fine immediately..  there are many activities which gets executed in the background during the startup..

If you restart the system without stopping the channels:

During the startup, all your sender channels gets activated and start polling thus maximum memory consumption which might impact your system(out of memory errors etc).

If we stop the channels, then we have an option to go for a control start... we can only start selective channels based on the backlog..

1. For all sender /receiver  system have these configuration either message will processed completely or roll back .

2. If Message will fail in PI we can reprocess that message

It depends on adapter type & error message that you get.. If it is picked then i don't think it'll rollback.

[Former Member · Accepted solution]

Thanks Hareesh for giving a new view point .

But in my current situation we are starting /stopping all channel at once and till now we didn't face any kind of error . as we try to down server at non business hours .

and yes If message is completely picked it will not roll back but partially message will roll back .

Thanks again

[former_member184720 · Accepted solution]

you are welcome

Yeah it works when you don't have incoming messages n got enough hardware resources to handle the startup activities.. after all SAP assured a 90sec restart for AEX

and for me sounds like you don't have many JMS/MDM channels..

[Former Member · Accepted solution]

correct ... no JMS/MDM channel

[engswee]

Hi Rudra

It is not mandatory, but it is the recommended practice - it is even listed on SAP Help Library

Starting and Stopping - Administering PI (Process Integration) - SAP Library

In step 3 of the the stop procedure:

Stop all sender communication channels on your central Advanced Adapter Engine to avoid new messages coming in. Otherwise, messages might still be picked up and lost if a point-in-time recovery of your database is required, for example, for patching, upgrade, or hardware switch.

Rgds

Eng Swee

[Former Member]

Dear Eng Swee,

I know its SAP recommended that's y we are doing this tedious task.

But If the complete server itself is down ,Central Advanced  Adapter Engine will be down  for sure , so how will new messages will come in PI .

and Second point if some patching , upgrade is going on I never saw that any message got lost since in that time also we have some messages inside PI server . so if we are not playing with database (particularly table) where messages are stored it will never lost .

Please let me know if I am wrong .

Regards,

Rudra

[Harish]

Hi Rudra,

Consider the scenario where your server downtime is 4:00 to 6:00. your sender channel pick the message at 3:59 and when message enter the PI the server will down. in this scenario data will be lost.

To prevent this we should stop the sender channel before the server downtime and make sure all the messages are processed before the downtime.

regards,

Harish

[Former Member]

Hi Harish,

If message is completely picked by channel and sever is down then it will be fail in adapter engine . or there is another point  if message is picked at 3.59 and we stopped channel at 4.00 , it is same as server downtime .

I wont see much difference in both cases .

Please let me know if i m wrong

Thanks & Regards

Rudra

[Harish]

Hi Rudra,

I believe the channel are stopped before the downtime (might be 10 min before).

It will not have much difference in case of ICO because of message processing on adapter engine, but in case of pipe line object it can have more impact. you might have different timing/maintains of adapter and integration engine. and if integration engine is down then there is possibility of message lost.

I agree that it is one odd case and having very less possibility but in case of production env. no one want to take the risk.

regards,

Harish

[Former Member]

Thank you so much Harish.

So I can assume there wont be much issue if we don't stop channels in PI but for since its a production system so we need to be extra careful .

Regards,

Rudra