ZNTA-939 Add support for wildfly 10

[seanf]

https://zanata.atlassian.net/browse/ZNTA-939

Also add logging filter configuration for some unavoidable startup warnings.

[seanf]

@zanata-jenkins retest this please

[seanf]

@zanata-jenkins retest this please

[seanf]

Tests passed on EAP 6; now rerunning on WildFly.

[carlosmunoz]

Seems to be all-config. Nothing to object.
👍

[carlosmunoz]

I take that back. Needs documentation on it now running on WF 10

[seanf]

Needs documentation on it now running on WF 10

Okay, but can I get you to update zanata-overlay (including standalone.xml) for WF 10?

[seanf]

@zanata-jenkins test this please

[seanf]

@zanata-jenkins test this please

[djansen-redhat]

Requires, but recommended? Or is "later than 6.4.6.GA" recommended?

[seanf]

Well, the idea is that EAP is recommended (6.4.6 or later), but WildFly 10
is also... possible.

On 16 March 2016 at 13:51, Damian Jansen notifications@github.com wrote:

In docs/release-notes.md
#990 (comment):

@@ -33,6 +33,9 @@

• ZNTA-928 - Readonly project doesn't have "lock" icon in UI
• ZNTA-793 - Remove unused method in TranslationMemoryAction

+

Infrastructure Changes

+* Zanata now requires JBoss EAP 6.4.6.GA or later (recommended), or WildFly version 10.x.

Requires, but recommended? Or is "later than 6.4.6.GA" recommended?

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
https://github.com/zanata/zanata-server/pull/990/files/5ebc7bcb5d17ad1f2421243ecd6379f78e6142fd#r56278787

Sean Flanigan

Principal Software Engineer
Globalisation Tools Engineering
Red Hat

[seanf]

@djansen-redhat See 5442e0e

[djansen-redhat]

It says it requires wildfly 10, but the appserver profile is wildfly8

[seanf]

It really means wildfly8+, because it was used for WildFly 8 and 9 as well.
It only affects developers, so we haven't bothered renaming it to 'wildfly'
yet. (Because changing it will knock-on to build scripts, CI jobs, etc.)

On 16 March 2016 at 14:05, Damian Jansen notifications@github.com wrote:

It says it requires wildfly 10, but the appserver profile is wildfly8

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
#990 (comment)

Sean Flanigan

Principal Software Engineer
Globalisation Tools Engineering
Red Hat

[djansen-redhat]

@seanf I'm getting this error when starting the server under Wildfly 10.

[seanf]

@seanf I'm getting this error when starting the server under Wildfly 10.

@djansen-redhat Did you try again on eap 6?

I might need you to email me the details of the database you were testing with, perhaps including the contents of the DBCHANGELOG table.

[djansen-redhat]

@seanf It appears to works fine on EAP6.4. Will email.

[seanf]

Looks like we have a problem with the standalone.xml in zanata-overlay for wildfly 10. I'll have to investigate.

[seanf]

@carlosmunoz comparing the old (wildfly 9) config with the wildfly 10 config in zanata-overlay, there seem to be quite a few changes.

Did you mean to add the system property document.storage.directory? Also, the document-storage-directory JNDI string has changed from ${user.home}/zanata/files to ${user.dir}/zanata/documents. From what I can tell, MigrateRawDocumentsToFileSystem uses the JNDI string if set, but tries to fall back on the system property otherwise. The system property is probably only intended for command-line invocations of Liquibase.

It seems like any existing WildFly installations may lose access to their uploaded documents, unless they move the directory contents around.

The logging format has changed (used to use UTC timestamps, now it will use local time).

The 'hibernate' cache-container seems to have lost its jndi-name. I thought we were referring to this in persistence.xml, so I'm not sure how it can work.

The JNDI string java:global/zanata/smtp/port was added, but I don't think it is used anywhere in Zanata these days.

Finally, the socket bindings have changed a lot. They seem to be exactly the same as used in functional tests, which means the ports are offset by 100, except smtp which is port 2552.

I'm thinking of reverting most if not all of the above changes, unless you object.

[carlosmunoz]

Did you mean to add the system property document.storage.directory?

I don't think I had anything to do with that.

Also, the document-storage-directory JNDI string has changed from ${user.home}/zanata/files to ${user.dir}/zanata/documents. From what I can tell, MigrateRawDocumentsToFileSystem uses the JNDI string if set, but tries to fall back on the system property otherwise. The system property is probably only intended for command-line invocations of Liquibase.

But isn't that just a settings change? All those using the previous value should not need to change it.

The 'hibernate' cache-container seems to have lost its jndi-name. I thought we were referring to this in persistence.xml, so I'm not sure how it can work.

This was probably a remnant of when we were using Infinispan for hibernate caching, which we don't do. You can ignore all modifications of the infinispan section for hibernate if you wish.

The JNDI string java:global/zanata/smtp/port was added, but I don't think it is used anywhere in Zanata these days.

I think that's obsolete now since we configure the smtp settings directly on the mail session in standalone.xml

Finally, the socket bindings have changed a lot. They seem to be exactly the same as used in functional tests, which means the ports are offset by 100, except smtp which is port 2552.

That doesn't make sense. They should be set to the default that come from wildfly.

I'm thinking of reverting most if not all of the above changes, unless you object.

I think I'm ok. I think the reason for this might have been basing the xml on a working functional test one, together with some unused and/or leftover settings.

[seanf]

It looks like the change in liquibase's checksum for changeset db/changelogs/db.changelog-3.1.xml::6::damason@redhat.com was caused by the creation of the system property document.storage.directory. This system property is referenced in db/changelogs/db.changelog-3.1.xml and passed as a CustomChange parameter. The parameters to a CustomChange are included in the checksum (by default), so the checksum changed.

I have removed the new system property and reverted a few other changes to make the wildfly 10 config closer to the previous wildfly 9 config.

[seanf]

On 18 March 2016 at 15:38, Carlos Munoz notifications@github.com wrote:

Did you mean to add the system property document.storage.directory?

I don't think I had anything to do with that.

Also, the document-storage-directory JNDI string has changed from
${user.home}/zanata/files to ${user.dir}/zanata/documents. From what I can
tell, MigrateRawDocumentsToFileSystem uses the JNDI string if set, but
tries to fall back on the system property otherwise. The system property is
probably only intended for command-line invocations of Liquibase.

But isn't that just a settings change? All those using the previous value
should not need to change it.

I'm assuming that end users will run the installer again, because with the
messaging subsystem changes, their old wildfly 9 config won't work. If they
generate new config with different settings, the behaviour of Zanata will
change.

The 'hibernate' cache-container seems to have lost its jndi-name. I
thought we were referring to this in persistence.xml, so I'm not sure how
it can work.

This was probably a remnant of when we were using Infinispan for hibernate
caching, which we don't do. You can ignore all modifications of the
infinispan section for hibernate if you wish.

The JNDI string java:global/zanata/smtp/port was added, but I don't think
it is used anywhere in Zanata these days.

I think that's obsolete now since we configure the smtp settings directly
on the mail session in standalone.xml

Finally, the socket bindings have changed a lot. They seem to be exactly
the same as used in functional tests, which means the ports are offset by
100, except smtp which is port 2552.

That doesn't make sense. They should be set to the default that come from
wildfly.

I'm thinking of reverting most if not all of the above changes, unless you
object.

I think I'm ok. I think the reason for this might have been basing the xml
on a working functional test one, together with some unused and/or leftover
settings.

True.

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
#990 (comment)

Sean Flanigan

Principal Software Engineer
Globalisation Tools Engineering
Red Hat

[seanf]

@carlosmunoz can you please take a look at 2daaba7 ?

[carlosmunoz]

2daaba7 reviewed 👍

[djansen-redhat]

✅ tested

[seanf]

Squash-merged as 5b2d8ca