May 2008 - jboss-dev-forums - Jboss List Archives

[Design of Messaging on JBoss (Messaging/JBoss)] - Re: Do we need JournalStorageManager::ADD_MESSAGE?

by clebert.suconic＠jboss.com

The Buffers were being created about 3 times for the AddMessage | MessagingBuffer buffer = new BufferWrapper(1024); // 1st | buffer.putByte(ADD_MESSAGE); | // this would | buffer.putBytes(message.encode().array()); //2nd | messageJournal.appendAddRecord(message.getMessageID(), buffer.array()); // 3rd | With this code, a class calling JournalStorageManager directly was performing about 30K inserts, with intermediate commits (what would require them to wait). I have made few changes: I have introduced an interface EncondingSupport: public interface EncodingSupport | { | int encodeSize(); | void encode(MessagingBuffer buffer); | } | And on the Journal it's possible to pass an Encodingsupport. With that it is possible to avoid at least few copies, writing to the final and necessary buffer only. (before calling journal.append). I could get twice about 60K writes/second. I know the transport now won't let me go above 8K / second based on NonPersistentMessages (on my computer at least)... but even that way I got some higher numbers on few tests I made. I'm still not ready to commit though... it should be another day :-) View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149572#4149572 Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4149572

17 years, 11 months

1
0
0 / 0

[Design of EJB 3.0] - Re: Nightly aggregator build is flawed

by heiko.braun＠jboss.com

Cool, thanks. I did a manual deploy from my machine today in order to sync with the latest WS test coverage. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149563#4149563 Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4149563

17 years, 11 months

1
0
0 / 0

[Design of JBoss Build System] - Re: Snapshots update (mvn -U)

by ALRubinger

See: http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149560 For underlying errors and possible resolution by Rajesh in QA. S, ALR View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149562#4149562 Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4149562

17 years, 11 months

1
0
0 / 0

[Design of EJB 3.0] - Re: Nightly aggregator build is flawed

by ALRubinger

"ALRubinger" wrote : The Clock on my Hudson machine, which also does a deploy, is good. Scratch that. My clock is NTP-synced, but only does install, no deploy. I don't think Carlo is deploying either. "rajesh" wrote : (03:14:31 PM) rajesh: ALR, : you are right. the culprit is dev65. | (03:14:31 PM) rajesh: [hudson@dev65 ~]$ date | (03:14:31 PM) rajesh: Tue May 6 22:24:34 EDT 2008 | (03:15:08 PM) rajesh: ALR, i will get this corrected. | (03:15:19 PM) rajesh: ALR, and check others too. So there's at least one violator found. :) S, ALR View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149560#4149560 Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4149560

17 years, 11 months

1
0
0 / 0

[Design of EJB 3.0] - Re: Nightly aggregator build is flawed

by ALRubinger

"rajesh" wrote : (02:44:42 PM) rajesh: ALR: looking at the forum thread, particularly the ejb3 snapshots, have you configured any hudson ejb3 jobs to deploy snapshots to repository? | (02:45:16 PM) rajesh: ALR: the hudson server time is in sync. But the build could have happened on any of the slave machines.. | (02:45:36 PM) rajesh: ALR, it would be easier to check if the job that is configured to deploy snapshots is known. The Clock on my Hudson machine, which also does a deploy, is good. S, ALR View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149558#4149558 Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4149558

17 years, 11 months

1
0
0 / 0

[Design of JBoss Build System] - Re: Snapshots update (mvn -U)

by heiko.braun＠jboss.com

Since my latest deploy you cannot see the wrong artifacts in the repository as they have been when I first published this post. They have been replaced. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149553#4149553 Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4149553

17 years, 11 months

1
0
0 / 0

[Design of JBoss Build System] - Re: Snapshots update (mvn -U)

by heiko.braun＠jboss.com

What's really strange is that when I deployed manually from my box, it did reuse the 61 qualifier and put the right timestamp in the filename. So basically it did overwrite an existing snapshot. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149551#4149551 Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4149551

17 years, 11 months

1
0
0 / 0

[Design of EJB 3.0] - Re: Nightly aggregator build is flawed

by heiko.braun＠jboss.com

This one http://hudson.qa.jboss.com/hudson/view/EJB3/job/EJB3_Aggregator_Deploy/ View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149547#4149547 Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4149547

17 years, 11 months

1
0
0 / 0

[Design of JBoss Build System] - Re: Snapshots update (mvn -U)

by heiko.braun＠jboss.com

I think the filename is build from the local system time. Which means by simply producing a build on box which not synced by NTP produces wrong artifacts. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149546#4149546 Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4149546

17 years, 11 months

1
0
0 / 0

[Design of Messaging on JBoss (Messaging/JBoss)] - Re: Do we need JournalStorageManager::ADD_MESSAGE?

by clebert.suconic＠jboss.com

Looking at JournalStorageManager, every time you use appendAddRecord or appendUpdateRecord (transactional or not), you are aways adding the RecordType to the Byte Array. Instead of using the ByteArray to store that information I would change the signatures on addRecord, addRecordTransaction, updateRecord, updateRecordTransaction. Something like: appendAddRecordTransactional(long txID, byte recordType, long id, byte[] record, IOCallback callback) appendUpdateRecordTransactional(long txID, byte recordType, long id, byte[] record, IOCallback callback) throws Exception; And same thing for non transactional. (Just for the record, for obvious reasons you don't need that on delete) I would then add the byte recordType to org.jboss.messaging.core.journal.RecordInfo. View the original post : http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4149522#4149522 Reply to the post : http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&p=4149522

17 years, 11 months

1
0
0 / 0

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

jboss-dev-forums May 2008