Steffen Larsen opened 1 decade ago
|
|
It seems to come from the tigase.xmpp.impl.Message.java class around line 181 (from master branch). Apparently conns (connected resources) are less than zero. Strange. |
|
Hi again, This issue is quite critical. Do you guys need some help or are you on it? |
|
I am sorry it's taking so long on our side. We are quite busy working through a queue of tasks and projects. If you can provide any help it would certainly speed up resolving it. |
|
Hi, I've run some tests on the latest version from repository and I wasn't able to replicate the issue, the log:
I've run the test on both AMP and message plugins.
|
|
Hi, Yes of course. I'll provide it tomorrow. The tigase-server version is latest 5.2. But I check exact number. Is there a way to check the version number of the war file? |
|
Thanks.
Neverhteless - it should be available in the manifest file inside war/jar |
|
ohh sorry, lost in java translation! :-) of course I meant the jar file. I'll check the manifest and give feedback tomorrow. |
|
As you can see from the image, a user (admin@yousee.dk) is connected twice. One with resource zool and the other with Psi. Sending a message fails. Version: 5.1.4 (in cluster mode) Init.properties: @config-type=--gen-config-def --admins=admin@p-xmpp01.xxx.idk, admin@xxx.dk --virt-hosts = xxx.dk --debug=server --user-db=mysql --user-db-uri=jdbc:mysql://localhost:3306/tigase?user=root&password=xxx --cluster-mode=true --cluster-nodes=p-xmpp01.xxx.idk,p-xmpp02.xxx.idk@ I've also tried it without the clusering enabled. And that actually made it work. So the problem does not lay in the versions of the tigase server, but more in the routing when clustering is enabled. :-( It actually also works in version 5.2.0 - beta3. Which I have tested here: Version: 5.2.0-beta3 (b3269) (stand alone mode). Init.properties: @config-type = --gen-config-def --virt-hosts = yousee.dk --cluster-mode = false --debug = server --user-db = mysql --admins = test2@yousee.dk --user-db-uri = jdbc:mysql://localhost/tigasedb?user=tigase&password=adminadmin@ |
|
I am afraid this is a known problem with clustering on version 5.1.x. This has been fixed in the last 5.2.0 version. |
|
Ups, I am sorry, I misread the original post. I am taking back what I said :-) please ignore my post. The problem may happen in a clustering mode as it is not well tested yet. Thank you for reporting it. This helps us with our tests. We will look at the problem during our tests. Could you please, just tell me what clustering mode you had enabled while experiencing the problem? Was it using ACS component or default clustering? |
|
Hey Artur! I am using the default cluster strategy. Maybe I should try out my scenario for 5.2.0-beta3 as well?. Will ACS work for 5.1.4? because I would very much like to try it out. Beside this message problem when in cluster mode, I am sometimes having trouble with login due to resource conflict, even though I am using different resources for my clients. Have you experienced this before? It only happens in cluster mode. -Cheers! /Steffen |
|
There is a different version of ACS for 5.1.4 although the ACS included in the 5.2.0 Beta3 should work with 5.1.4. As for the resource conflict, there were problems like this but they were fixed in subsequent versions. Hard to tell about your case though without investigation. |
|
Tested with latest version with and w/o ACS hence it just struck me that this might be related to the changes in delivery logic (see linked changeset and tickets) |
|
Wojciech, I suppose you confirm that this happened during your tests as well? |
|
yes, issue is reproducible in the cluster environment with both available strategies. |
|
So to summarize: the direct presence, sending messages to the same user with different resource, does not work in any clustered scenario (default strategy or ACS) ? and not in 5.1.x or in 5.2? .Cheers! /Steffen |
|
Any news in this issue? |
|
Yes, I have news. I am working on the issue right now. I have covered this case in our cluster automated tests and I can replicate and confirm this is really a problem. I am working on a fix. I expect to have it fixed today or tomorrow. |
|
Super cool Artur!. I believe that it will be both on the 5.1.x and 5.2.x branch? Just say when it can be tested, I have a full testbed here as well. :-) |
|
I am working on 5.2.0 release right now, so for sure it will work on this version. 5.1.0 may be fixed as well depending how much work it needs. |
|
Hi Artur, I can only see a few changes that might have something to do with this issue on master. I can see some NPE fixed on https://projects.tigase.org/projects/tigase-server/repository/revisions/f65f1d25375e055d886ee15866bdcc374b56d20e - but I am not sure that this is related to this issue here. :-) I there is a fix on master, then let me know so I can test it. :-) Cheers and thanks Artur! /Steffen |
|
The fixes were done in SessionManager, thats for sure. They could be accompanied by some code changes in connection managers as well. So you should really start investigation from SM changes tracking. I am very sorry I cannot provide you with more details. Unfortunately these were kind of fast/hot fixes without a ticket report attached to code changes. So I do not even remember when exactly they were made. Andrzej was the developer who was making the fixes. Let's ask him where to look and when this could be done. Andrzej: could you please provide any details if you have them? |
|
Ups sorry everybody. I am responding to so many queries and some of them are very similar, so I messed this up. Again! Please forget my previous comment. Stephen the cluster fix is not available in repository yet. I am still working on it. Instead of a quick dirty fix I found a way to do it in a better way, more reliable but it needs some work, therefore it will take me a few more days. I am sorry for keeping you waiting. |
|
Hi Artur, Thats ok. Just ping me via. IM or mail, when there is a fix on master. I'll synch my source and test it straight after that. /Steffen |
|
Andrzej, you have made some fixes in this part of the code recently. Could you please confirm that the issue is fixed? |
|
I made a fix in delivery of presence stanzas to other resources of same user and it is fixed now, so delivery of direct presence to other resources works fine. As for delivery of messages to other resources on diffrent nodes, there is still an issue there which needs to be fixed. |
|
OK cool. So it means that direct messages to full JID now works on single instances of Tigase, but not in a cluster? |
|
Applied in changeset tigase-server|commit:c86055dfb29cd582b7b698d2ae8db658ffe64538. I applied a fix for delivery of messages sent to resources connected to other cluster nodes. |
|
Super!. I am looking forward to try out the fix. Will try it tomorrow or the following days. Thanks! |
|
Referenced from commit 1 year ago
|
Type |
Bug
|
Priority |
Major
|
Assignee | |
RedmineID |
1480
|
Version |
tigase-server-5.2.0
|
I am trying to send from the same user with one resource to anohter resource.
That is: sending a message stanza from test2@test.dk/Phone to test2@test.dk/Web. This have worked before when using tigase 5.1 (using 5.2 now), so I wonder what have changed?.
Of course it could be AMP or Message module that have changed, but I really cant see it.
Here are the stanzas (from test2@test.dk/Phone client):
@
jkhjkhk jkhjkhkThe recipient is no longer available.
@
My init.prop (sm-plugins) looks like this:
Ive tried using both AMP and message as sm-plugins.