2

I have recently deployed WSO2 API Manager (2.0.0) as 2 instance all-in-one clustered (using Hazelcast AWS scheme) with mysql datasource as specified in this link

Since, not able to find a complete step by step installation guide for this setup. I would like to clarify few areas that I am not too sure of.

  1. Depsync via SVN - since this will be manger to manger nodes (instead of manager to worker nodes) both will have <AutoCommit>true</AutoCommit>. Should we have any concern on this?
  2. DAS - Having DAS as separate node, should both WSO2AM and WSO2DAS share the same WSO2AM_STATS_DB database?
  3. Publisher - Can we use both publishers (i.e one at a time). Noticed once we published an API, it takes time for other publisher to sync the state to published (even if the new API appears almost immediate on other publisher as created)

Thank you.

Ravi N.
  • 81
  • 2
  • 10

1 Answers1

1

1) If you enable <AutoCommit>true</AutoCommit> in both nodes, it can cause svn conflicts if there are parallel publishing from 2 nodes. Instead, you can publish to multiple gateways from the publisher. For that, you can configure multiple environments in <Environments> section in api-manager.xml

2) Yes, DAS writes summarized data to that DB, and APIM dashboards read data from the same DB.

3) All publisher/store nodes should be in the same cluster. Then only they can communicate about API state changes etc. To be on the same cluster, all these nodes should have the same clustering domain. You can configure that in clustering section of axis2.xml.

Bee
  • 12,251
  • 11
  • 46
  • 73
  • Couple followup questions 1) I am on AWS and nodes may get auto deployed. Maintaining IP will be hard. Therefore, I prefer avoid publisher's multiple gateway environment configuration. When you mentioned may cause conflicts due parallel publishing, this I assume when admin decided to add or modify API in publisher? With proper control on access to publishers - we able to mitigate this issue? 3) Each node runs as single JVM (without profile) - Publisher/Store/Key Manager/Traffic Manager/GW in one JVM and clustered. Am I correct to say Publisher/Store in this case are clustered? – Ravi N. Oct 28 '16 at 03:16
  • "Maintaining IP will be hard. " You can use hostnames. 1) if you can guarantee such concurrent api creation/modification won't happen, svn model should work. or you can use rsync. 3) yes, if both pub and store are in same jvm, they can be considered clustered. But what I meant was that you 2 nodes need to be clustered so that when 1 publisher changes the status of an api the 2nd publisher can know it. – Bee Oct 29 '16 at 15:41