Description of problem: There should be a step that describes execution of tendrl-ansible command in upgrade guide [1] as part of solution for bz 1650557. tendrl-ansible should be probably called after `# yum update` is called on WA server or at the end of upgrade. There should be also a link to section with tendrl-ansible and mention how to run it. [1] https://access.qa.redhat.com/documentation/en-us/red_hat_gluster_storage/3.4/html-single/quick_start_guide/#red_hat_gluster_storage_web_administration_3_4_x_to_3_4_y
It would be better to get ack from dev team, but my opinion is, that the best place for this additional step is between step 3 (Run the tendrl-upgrade script) and step 4 (If updates to the kernel package occurred, reboot the server system. If not, restart the Web Administration services). And the new step should look somehow like this: Run tendrl-ansible playbooks with the same configuration as for initial installation. For more information see section 3.3. Web Administration Installation[1], steps 8 and 9. [1] https://access.qa.redhat.com/documentation/en-us/red_hat_gluster_storage/3.4/html-single/quick_start_guide/#web_administration_installation
One small change for the new step: Run tendrl-ansible playbooks with the same configuration as for initial installation. For more information see section 3.3. Web Administration Installation[1], steps 2 to 9.
@Daniel, Running tendrl-ansible can be the last step. Also note that we can remove the step of restarting the tendrl services(barring httpd and collectd) as the tendrl-ansible will take care of the same. But I don't think httpd and collecd are restarted by tendrl-ansible. What do you think?
It might be possible to drop the steps for restarting Tendrl related services in the future and let tendrl-ansible do that task completely. But for now I would rather keep it as it is - manually instruct user to restart all the related services at the end of the update process. The reasons are follows: * tendrl-ansible ensure, that the required services are running, they are not explicitly restarted (except following case) * tendrl-ansible restart only those services where configuration was changed by tendrl-ansible * tendrl-ansible doesn't manage tendrl-gluster-integration and collectd on Gluster Storage Servers If we would like to drop the steps with manual restart of Tendrl services, we have to extensively test multiple related scenarios to ensure, that there is no possibility, that some of the services will get to unexpected state. I think, we shouldn't take the risk of breaking something so late in the schedule of this release. Restarting of all the related services at the end of the update process should be safe, without any side effect. But for the future releases, it is worth to consider enhancement of tendrl-ansible to take care of some or all of the additional update steps.
To sum this up: because of BZ 1650557, we need to enhance our documentation related to update procedure. In section "4.2. Red Hat Gluster Storage Web Administration 3.4.x to 3.4.y", we need add a step like this (drafted in comment 2 and comment 3): > Run tendrl-ansible playbooks with the same configuration as for initial > installation. For more information see section 3.3. Web Administration > Installation[1], steps 2 to 9. > > [1] https://access.qa.redhat.com/documentation/en-us/red_hat_gluster_storage/3.4/html-single/quick_start_guide/#web_administration_installation Now, there are 2 places where we could add it (as Daniel and Nishanth point out): a) after step #2 (yum update) and before step #3 (tendrl-upgrade) b) after step #3 (tendrl-upgrade) and before step #4 (restarting services) c) after step #4 (restarting services) Option a is original proposal from Filip, option b is suggested by Daniel in comment 2, option c by Nishanth in comment 4. I'm asking Nishanth to verify this description, and indicate if he agrees with Daniel to go with option b or would like to go with c. ## Additional Details, not important for the work which needs to be done here (doc team can safely ignore this section) As Daniel points out, there is no way to delegate service restart into tendrl ansible now, as that would require changes in tendrl-ansible. The best way would actually be to: * move installation of gluster integration from import job to tendrl-ansible (eg. https://github.com/Tendrl/tendrl-ansible/issues/126) * actually care about the update and allocate the dev effort for it, which includes update playbook But this is outside of scope of this bugzilla or BU2.
Created attachment 1511377 [details] screenshot with documentation, highlighting options where to add the additional step I'm attaching a screenshot with documentation, highlighting options where to add the additional step.
I will suggest running tendrl-ansible between 3 and 4. In initial steps, we are suggesting to stop all tendrl related services So I feel running ansible between 3 and 4 will definitely start the services again. (already httpd also started by tendrl-ansible). and in step 4 we have to consider two scenarios: 1. managed cluster: if the cluster is managed then we have to worry about starting gluster-integration and collectd services. So we will suggest step only for the already managed clusters. 2. Unmanaged cluster: if the cluster is not managed yet but if the user doing the update then step 4 is not required, because we are going to run tendrl-ansible. If they run collectd and tendrl-ansible as per that downstream document then import flow will be screwed up. (if it is a fresh machine then it is fine, but if it is already unmanaged cluster and ready to import then it is a problem, they have to unmanage again to solve this problem).
As per Daniels comment if tendrl-ansible not start all stopped services then we have to modify step 4 like what are all the services should be started for managed cluster and what are all the services required to start for an unmanaged cluster.
Agree to go forward with option (b) now. (In reply to Martin Bukatovic from comment #6) > To sum this up: because of BZ 1650557, we need to enhance our documentation > related to update procedure. > > In section "4.2. Red Hat Gluster Storage Web Administration 3.4.x to 3.4.y", > we need add a step like this (drafted in comment 2 and comment 3): > > > Run tendrl-ansible playbooks with the same configuration as for initial > > installation. For more information see section 3.3. Web Administration > > Installation[1], steps 2 to 9. > > > > [1] https://access.qa.redhat.com/documentation/en-us/red_hat_gluster_storage/3.4/html-single/quick_start_guide/#web_administration_installation > > Now, there are 2 places where we could add it (as Daniel and Nishanth point > out): > > a) after step #2 (yum update) and before step #3 (tendrl-upgrade) > b) after step #3 (tendrl-upgrade) and before step #4 (restarting services) > c) after step #4 (restarting services) > > Option a is original proposal from Filip, option b is suggested by Daniel in > comment 2, option c by Nishanth in comment 4. > > I'm asking Nishanth to verify this description, and indicate if he agrees > with > Daniel to go with option b or would like to go with c. > > ## Additional Details, not important for the work which needs to be done here > > (doc team can safely ignore this section) > > As Daniel points out, there is no way to delegate service restart into tendrl > ansible now, as that would require changes in tendrl-ansible. > > The best way would actually be to: > > * move installation of gluster integration from import job to tendrl-ansible > (eg. https://github.com/Tendrl/tendrl-ansible/issues/126) > * actually care about the update and allocate the dev effort for it, which > includes update playbook > > But this is outside of scope of this bugzilla or BU2.
Looks good. --> VERIFIED