Is there any startup / shutdown order in a cluster?
I've never heard that there is any startup or shutdown order in PRPC but I have found this article that says Search node must be started first (https://community.pega.com/knowledgebase/articles/performance/best-practices-starting-and-stopping-nodes-pega-cluster). It doesn't talk about Stream and it is confusing. If this is the case, such important process must be stated in the installation guide or help overly with much more emphasis. For customer who maintains many nodes, it is pretty big deal as for one node to start or shutdown takes minutes and they have to wait for 10 - 20 minutes in series, every time. The best scenario for them is no order, and Pega cluster takes care of node dependencies so they can run shutdown script for all nodes at the same time. If not, it has to be explained why, and what would be the outcomes if they don't follow it.
Let's say, if I have below 8 node types, in which order should I start or shutdown? Or does it really matter? Also please let me know until when should I wait to start / shutdown the next nodes (ex. until certain step is done, or row in a pr_sys_statusnodes table status becomes certain status like "Shutdown" or deleted, etc).
I've never heard that there is any startup or shutdown order in PRPC but I have found this article that says Search node must be started first (https://community.pega.com/knowledgebase/articles/performance/best-practices-starting-and-stopping-nodes-pega-cluster). It doesn't talk about Stream and it is confusing. If this is the case, such important process must be stated in the installation guide or help overly with much more emphasis. For customer who maintains many nodes, it is pretty big deal as for one node to start or shutdown takes minutes and they have to wait for 10 - 20 minutes in series, every time. The best scenario for them is no order, and Pega cluster takes care of node dependencies so they can run shutdown script for all nodes at the same time. If not, it has to be explained why, and what would be the outcomes if they don't follow it.
Let's say, if I have below 8 node types, in which order should I start or shutdown? Or does it really matter? Also please let me know until when should I wait to start / shutdown the next nodes (ex. until certain step is done, or row in a pr_sys_statusnodes table status becomes certain status like "Shutdown" or deleted, etc).
- WebUser
- WebUser
- BackgroundProcessing
- BackgroundProcessing
- Search
- Search
- Stream
- Stream
Usually I would put Search node or Stream node with BackgroundProcessing node together but here in this example I have separated it so things are clear.
Thanks,
***Edited by Moderator Marissa to update Platform Capability tags****
Pegasystems Inc.
US
Since you are running Pega 8.4 (since Pega 8.3+), you should not have to follow specific order except that you should start stream node one by one first due to a known issue. The rest node types you can start in any order.
IT Solution Service
SG
Thanks. I understood this way - Pega cluster is designed such way that any node can get started regardless of its node type but in 8.3 and 8.4, Stream has to be started one by one first due to some technical issue.
May I know what kind of issue, or BUG number? I want to explain that to customer so they know the impact when they didn't follow it.
Delta Dental of California
US
@KevinZheng_GCS : is there any resolution planned for this known issue and remove the dependency on the order of restart? If so, in which version is it planned to be made available?
Updated: 29 Oct 2020 6:11 EDT
Credera
US
Hello!
I have these node groupings and was told by a Pega infrastructure specialist that you should start/shutdown in the following order:
Bring Up:
- Background Node 1 (BackgroundProcessing, Search, Stream, WebUser)
- use the Web UI to verify the search index has finished building before continuing. I have not found a way to programmatically confirm search is finished building
- Background Node 2 (BackgroundProcessing, Search, Stream)
- Decisioning Node 1 (DDS, ADM, RTDG, Batch, RealTime)
- You must have port 7000 open for Cassandra but it is not in the documentation
- Decisioning Node 2 (DDS, ADM, RTDG, Batch, RealTime)
- Decisioning Node 3 (DDS, ADM, RTDG, Batch, RealTime)
- Web Node 1 (WebUser)
- Web Node 2 (WebUser)
Bring down in the opposite order.
There are some errors that happen, will link to the community post I created.
W&W Informatik GmbH
DE
@ShelbyS6 That's not the answer to the question. you describe the case with three nodes: Background, Decisioning and WebNode. What about: If you have separate nodes: Background, stream, index and webnodes?
Updated: 25 Oct 2021 1:57 EDT
Siemens
@CloeW938 Yes the same issue we were faced in version 8.1.X and Pega suggested the same. The search node should be started first.
Node A - DNodeType=WebUser " Node B - DNodeType=BackgroundProcessing,Stream,WebUser,Search "
Need to Start Node B and then Node A. Need to Stop Node A and then Node B.