Bug 1158581 - Update the product JGroups files based on the values recommended in BZ1122562
Summary: Update the product JGroups files based on the values recommended in BZ1122562
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: JBoss Data Grid 6
Classification: JBoss
Component: JGroups
Version: 6.3.1
Hardware: Unspecified
OS: Unspecified
unspecified
unspecified
Target Milestone: ER1
: 6.4.1,6.5.0
Assignee: Alan Field
QA Contact: Martin Gencur
URL:
Whiteboard:
Depends On:
Blocks: jdg64-GA-Blockers
TreeView+ depends on / blocked
 
Reported: 2014-10-29 16:54 UTC by Alan Field
Modified: 2015-04-02 12:13 UTC (History)
10 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed: 2015-04-02 12:13:57 UTC
Type: Bug
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Red Hat Bugzilla 1122562 0 urgent CLOSED No documentation around JGroups settings 2021-02-22 00:41:40 UTC

Description Alan Field 2014-10-29 16:54:57 UTC
Description of problem:

The default JGroups configuration files included with JDG need to be updated to reflect the values suggested in the BZ1122562.

QE is already using these JGroups configuration files:
https://svn.devel.redhat.com/repos/jboss-qa/load-testing/etc/jdg-60/jgroups/infinispan60/jgroups-tcp.xml
https://svn.devel.redhat.com/repos/jboss-qa/load-testing/etc/jdg-60/jgroups/infinispan60/jgroups-udp.xml

Version-Release number of selected component (if applicable):


How reproducible:


Steps to Reproduce:
1.
2.
3.

Actual results:


Expected results:


Additional info:

Comment 2 Shay Matasaro 2014-10-30 13:44:03 UTC
 combining the information from some of my other cases , and conversations with numerous engineers, here is my executive summary:

rejection_policy should always be set to "Discard"

for the thread pools :
OOB
- Thread numbers min should be a minimum of 20
- thread number  max  should be a  high number which should be based on load 
-  queue should be disabled

 
Regular thread pool
- For the regular thread pool (non-OOB), min number of thread should equal to number of nodes
- max number should be slightly larger , so for a small grid(2-10 nodes) the ratio could be  2 times number of nodes ,  but for a a large grid (20 or more nodes) the ratio should be  smaller for example:
-- max 25 threads on a 20 node grid
-- max 110 threads on a 100 node grid.
- queue should be enabled  and the max default of 100 is fine



Unless needed SAs or customers should probably not tune other setting unless there is a very specific reason

Comment 5 Sebastian Łaskawiec 2015-01-15 14:48:16 UTC
PR: https://github.com/infinispan/jdg/pull/440

Comment 7 Alan Field 2015-01-16 15:53:25 UTC
Missed two timeout values in *PING protocols in these files:

server/integration/jgroups/src/main/resources/jgroups-defaults.xml
core/src/main/resources/default-configs/default-jgroups-tcp.xml

These values are deprecated and will be ignored by JGroups, but will cause log messages to be written:

09:16:34,666 WARN  [org.jgroups.stack.Configurator] (sc-main) JGRP000014: Discovery.timeout has been deprecated: GMS.join_timeout should be used instead

IMO, this isn't enough to trigger a CR3, but I'll send another PR with the changes

Comment 10 Alan Field 2015-01-21 16:38:35 UTC
In the JDG 6.4 CR3 build, the protocols used in the server stacks are not the same as the ones used in the embedded configuration files. Here is a PR to fix that:

https://github.com/infinispan/jdg/pull/459

Comment 11 Alan Field 2015-01-22 14:36:03 UTC
Verified with JDG 6.4 CR4

Comment 13 Sebastian Łaskawiec 2015-02-13 13:27:06 UTC
PR: https://github.com/infinispan/jdg/pull/494

Comment 14 Alan Field 2015-03-12 13:59:05 UTC
Verified in JDG 6.4.1 ER1


Note You need to log in before you can comment on or make changes to this bug.