Red Hat Bugzilla – Bug 847424
rhc tools reports failure for creating scaled java app but it succeeds
Last modified: 2012-08-23 16:38:13 EDT
Description of problem:
When creating a scalable jbossas-7 app the rhc command line tools reports failure
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. export pass="your openshift password"
2. rhc-create-app -a testscale1 -t jbosseap-6.0 -p $pass -s -d
3. watch error
Creating application: testscale1 in whearn
Problem reported from server. Response code was .
Failed to access resource:
It to succeed and check out the git repository.
Forgot to post the success line from user_action.log:
SUCCESS DATE=2012-08-10 TIME=22:31:48 ACTION=ADD_APPLICATION REQ_ID=e6c5c77932bb4467af2f81af5d8085a9 USER_ID=da8d582bbfc6480bafea382caf466f69 LOGINfirstname.lastname@example.org Created application testscale1
This is likely timeout based - JbossAS7 is the most time consuming cartridge to configure.
*** Bug 849313 has been marked as a duplicate of this bug. ***
Worked on this with Rajat. We were able to reproduce this using a direct request against the production broker using a curl command. This suggested that the problem was not in the rhc client tools.
Rajat and Ram ran with this and found the source of the problem: we are using Amazon's Elastic Load Balancer (ELB) with our production broker, and ELB has a 60-second timeout: https://forums.aws.amazon.com/thread.jspa?threadID=33427&start=50&tstart=50
In the long term, the broker and rhc client tools should be modified to handle the app creation process with multiple discrete actions:
1. User uses rhc client tools to request a new app
2. Broker responds with a ticket and kicks off the app creation
3. User can 'check in' with the ticket to monitor app creation progress
4. When the app is ready, user can use the ticket to set up the git clone
(User stories to this effect already exist in our backlog)
This approach makes an issue like a 60-second timeout unimportant to the implementation, and will ensure that the process works for ever-larger app setups.
In the meantime, we need to find another way to work around ELB's 60-second limit.
Reclassified and assigned to twoodson.
(In reply to comment #5)
> Reclassified and assigned to twoodson.
sorry; that's mwoodson.
I have disabled the AWS ELB. openshift.redhat.com is now round robin between our two proxy nodes.
This will fix the ELB timeout problem.
As per our end of print scrum, marking this fixed. Closing.