Using Trilio in a model with 2 Ceph clusters not working

Bug #1986650 reported by Meenu Sachan
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
TrilioVault Data Mover Charm
Triaged
Medium
Unassigned

Bug Description

We are facing an issue where backups are randomly failing. we discovered the issue lies in the Juju model which has two Ceph clusters (one for HDD, other for SDD).

Both of Ceph clusters are integrated with Cinder (by using multiple backends). This works just fine with Nova, because cinder will connect to both clusters and provide block volumes to the hypervisors with no issue.

The problem is that Trilio is capable of handling only one cluster, and this cluster is configured using the relation:
ceph-mon:client trilio-data-mover:ceph

This relation creates a ceph configuration file inside the nova-computes, named "/var/lib/charm/trilio-data-mover/ceph.conf", which has connection details to the ceph cluster. It also creates the necessary keyrings.

Then, the backup process uses the rbd client using this configuration file and keyring.

The result is the following:

1. VMs with disks only on the original Ceph cluster: OK

2. VMs with disks in the second Ceph cluster: NOK

Revision history for this message
Liam Young (gnuoy) wrote :

If trilio can support multiple ceph backends then this probably isn't that big a charm change

Changed in charm-trilio-data-mover:
status: New → Triaged
importance: Undecided → Medium
Revision history for this message
Abhijeet Patra (abhijeet-trilio) wrote :

Hi Liam Young,

I am attaching the steps/script which we used in our lab to setup the second cluster and make Trilio configure and work with it.

This has some modifications on the ceph clusters and also in the trilio-datamover nodes

https://pastebin.ubuntu.com/p/2MT7msvZS7

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.