[PATCH] 414 Direct deployment for TPS.
The deployment tool has been modified to deploy TPS directly from the share folder. This way the TPS UI can be upgraded automatically with RPM upgrade without having to write upgrade scripts. For this to work, the TPS web application files cannot contain any customization. So, the cfgPath parameter has been removed from web.xml, and the CMSStartServlet has been modified such that if the parameter is missing it would generate a default path matching the original value in web.xml. Also, the velocity.properties has been modified to use a fixed value for the file.resource.loader.path parameter pointing to the share folder. In the future other subsystems may be modified to use the same deployment mechanism. Ticket #748, #752, #499 -- Endi S. Dewata
New patch attached. Fixed a problem creating symlinks. -- Endi S. Dewata On 2/25/2014 2:24 PM, Endi Sukma Dewata wrote:
The deployment tool has been modified to deploy TPS directly from the share folder. This way the TPS UI can be upgraded automatically with RPM upgrade without having to write upgrade scripts.
For this to work, the TPS web application files cannot contain any customization. So, the cfgPath parameter has been removed from web.xml, and the CMSStartServlet has been modified such that if the parameter is missing it would generate a default path matching the original value in web.xml. Also, the velocity.properties has been modified to use a fixed value for the file.resource.loader.path parameter pointing to the share folder.
In the future other subsystems may be modified to use the same deployment mechanism.
Ticket #748, #752, #499
Hi Endi, First of all, thanks for your patience on irc this morning. As you know, the pkispawn/installation area is new to me, so I try my best to understand what the issues are and what you are trying to fix. Here is a recap of some of my observations and suggestions. None are critical that need to be "fixed", but rather suggestions to add clarification. * MSStartServlet.java - Since it took me some time to investigate and figure out what "context", "subsystem", and "instanceDir" came from and resolved to. It might be a good idea to add comments right above the section explaining how they come to. - About removing destroy(), I am sure you have good reasons, but if you must "remove" it, may I suggest you comment it out and put explanation in comment there for my consumption later? Again, this part just doesn't seem relevant to the ticket, and I'd really like to focus on our tickets at hand. If we keep finding different things to go off and pursue, we will have a hard time reaching our immediate targets. In general, if it's unrelated, I prefer filing it off and attending to it later. Or, if others are finding themselves having cycle to look at this part for you, I'll let them do the research and ack for you. I wash my hands at this time ;-). * slot_substitution.py - This took you quite some time to explain to me too ;-). I think since the new TPS web.xml and velocity.properties do not contain any "slots", whether you skip the substitution or not doesn't make that much difference, so adding the code to single out TPS is ok, however, may I suggest that you add a little comment there explaining that there will be no need for substitution because slots are not allowed in the /user/share or something to that effect? I just worry later when we need to get back to this we don't remember why (I'm sure you will, but maybe for my sake). * I also asked if this patch will affect stand-alone tomcat-TPS and you said no. It's basically an ACK if it tested to work. Again, adding clarity in comments are just suggestions. thanks, Christina On 03/05/2014 11:15 AM, Endi Sukma Dewata wrote:
New patch attached. Fixed a problem creating symlinks.
_______________________________________________ Pki-devel mailing list Pki-devel@redhat.com https://www.redhat.com/mailman/listinfo/pki-devel
On 3/7/2014 5:22 PM, Christina Fu wrote:
Hi Endi,
First of all, thanks for your patience on irc this morning. As you know, the pkispawn/installation area is new to me, so I try my best to understand what the issues are and what you are trying to fix.
Thanks for your review. I think this is a good exercise to familiarize (or re-familiarize) ourselves with various aspects of Tomcat and the deployment process which indeed is quite complex. This patch is part of an effort to simplify the process, as you saw by the reduction of the deployment code for TPS (and eventually for other subsystems too).
Here is a recap of some of my observations and suggestions. None are critical that need to be "fixed", but rather suggestions to add clarification.
* MSStartServlet.java - Since it took me some time to investigate and figure out what "context", "subsystem", and "instanceDir" came from and resolved to. It might be a good idea to add comments right above the section explaining how they come to.
I added some comments in the updated patch.
- About removing destroy(), I am sure you have good reasons, but if you must "remove" it, may I suggest you comment it out and put explanation in comment there for my consumption later? Again, this part just doesn't seem relevant to the ticket, and I'd really like to focus on our tickets at hand. If we keep finding different things to go off and pursue, we will have a hard time reaching our immediate targets. In general, if it's unrelated, I prefer filing it off and attending to it later. Or, if others are finding themselves having cycle to look at this part for you, I'll let them do the research and ack for you. I wash my hands at this time ;-).
I originally thought this was a straight forward clean up. I've created a separate ticket for this: https://fedorahosted.org/pki/ticket/896 It has a link to the source code that we discussed.
* slot_substitution.py - This took you quite some time to explain to me too ;-). I think since the new TPS web.xml and velocity.properties do not contain any "slots", whether you skip the substitution or not doesn't make that much difference, so adding the code to single out TPS is ok, however, may I suggest that you add a little comment there explaining that there will be no need for substitution because slots are not allowed in the /user/share or something to that effect? I just worry later when we need to get back to this we don't remember why (I'm sure you will, but maybe for my sake).
It's not just unnecessary, but the pki_target_subsystem_web_xml and pki_target_velocity_properties now point to non-existent paths because we're no longer copy the TPS files into those locations in the instance, so this code cannot be executed for TPS. I've added some comments in the code. if deployer.master_dict['pki_subsystem'] != "TPS": deployer.file.apply_slot_substitution( deployer.master_dict['pki_target_velocity_properties']) deployer.file.apply_slot_substitution( deployer.master_dict['pki_target_subsystem_web_xml'])
* I also asked if this patch will affect stand-alone tomcat-TPS and you said no.
Since currently stand-alone TPS for Tomcat doesn't work we cannot really verify this. But so far I don't see anything specific that would prevent that from working.
It's basically an ACK if it tested to work. Again, adding clarity in comments are just suggestions.
Yes, I have installed TPS and tested it with the CLI and UI.
thanks, Christina
Thanks. Pushed to master. -- Endi S. Dewata
participants (2)
-
Christina Fu -
Endi Sukma Dewata