Implement retries for MongoDB shell commands

[iberezovskiy]

Retries for MongoDB shell commands could be useful
in some cases:

• unstable connectivity
• when replica or sharding initialization is in process and
  it needs to wait some time till its end

[EmilienM]

I know this is too late, but it would have been great to have unit & functional testing.
For history, we are investigating failures that could probably come from this PR.

https://bugs.launchpad.net/tripleo/+bug/1488531

[iberezovskiy]

Emilien,

I think I can provide solution for this but it can look a little bit
revolutionary: if auth in mongo is enabled and we are going to create
replica set, we should follow these steps:

1. Prepare configuration file with admin user credeantial on each mongo
   node.
2. Deploy replica master. Initiate replica from it. (as result we will get
   master of replica on this node)
3. Create admin user on master (username and password should be the same as
   in credentials file)
4. Deploy another slaves. These slave will be able to add yourself to
   replica set using authentication creadentials for primary mongo.

what do you think?

P.S. this solution is used in Fuel deployments and I can prepare (and I
will do) the same approach for puppetlabs-mongodb.

2015-08-26 17:08 GMT+03:00 Emilien Macchi notifications@github.com:

I know this is too late, but it would have been great to have unit &
functional testing.
For history, we are investigating failures that could probably come
from this PR.

https://bugs.launchpad.net/tripleo/+bug/1488531

—
Reply to this email directly or view it on GitHub
#217 (comment)
.

Thanks, Ivan Berezovskiy
Ceilometer Deployment Engineer
at Mirantis https://www.mirantis.com/

slack: iberezovskiy
skype: bouhforever
phone: + 7-960-343-42-46

[dprince]

@iberezovskiy hi, the new proposal sounds interesting. What we are dealing with here is an interface breakage however and I'm inclined to say reverting the (breaking) code is the best option today.

Then perhaps we can post a new patch with the more "revolutionary" changes that you are proposing?

[iberezovskiy]

Dan Prince,

If it broke something I am agree to revert this change, but please, do not
revert commit with adding retries for mongo shell commands. They are useful
in any case :)
So, you can mark like an action items on me to prepare my patch, and what I
need here is review of my improvement from you and OpenStack Puppet team.

2015-08-26 17:55 GMT+03:00 Dan Prince notifications@github.com:

@iberezovskiy https://github.com/iberezovskiy hi, the new proposal
sounds interesting. What we are dealing with here is an interface breakage
however and I'm inclined to say reverting the (breaking) code is the best
option today.

Then perhaps we can post a new patch with the more "revolutionary" changes
that you are proposing?

—
Reply to this email directly or view it on GitHub
#217 (comment)
.

Thanks, Ivan Berezovskiy
Ceilometer Deployment Engineer
at Mirantis https://www.mirantis.com/

slack: iberezovskiy
skype: bouhforever
phone: + 7-960-343-42-46

[nibalizer]

According to http://ci.puppet.community/modules/puppetlabs/puppetlabs-mongodb, we haven't had passing beaker tests on this module, ever. @EmilienM pcci runs beaker tests per pull request and can be configured to vote here. However, no one has made the effort to make the tests pass.

[iberezovskiy]

Hi, Guys!
Can you please review this pull request #194 ?
Here is what I've been talking before: improvements for replica, for authentication. It follows all steps that I've listed before.

Thanks!