Adding initiator validation for optimised and inaccessible paths #4468
Conversation
Signed-off-by: manasagowri <[email protected]>
manasagowri
requested review from
Manohar-Murthy and
HaruChebrolu
as code owners
There was a problem hiding this comment.
Hi @manasagowri ,
Thanks for adding up new validation at initiator level. There are few comments on the code changes.
Lets discuss more over these comments.
| @@ -692,6 +692,10 @@ def failover(self, gateway, fail_tool): | |||
| LOG.info( | |||
| f"{list(active[0])} is new and only Active GW for failed {hostname}" | |||
| ) | |||
| active_gw = list(active[0]) | |||
There was a problem hiding this comment.
- Firstly, lets add client namespace validation in
ha.run. Lets not touch main definition. - Return back the active gateway node which is incharge of failed over GW in dict.
| @@ -896,6 +905,26 @@ def validate_incremetal_io(write_samples): | |||
|
|
|||
| LOG.info("IO Validation is Successfull on all RBD images..") | |||
|
|
|||
| def validate_initiator(self, gateway, namespaces, ana_id): | |||
There was a problem hiding this comment.
lets call this from ha.run()
| fail_gw_ana_ids.append(gw.ana_group_id) | ||
| ns = self.fetch_namespaces(gw, [gw.ana_group_id]) | ||
| namespaces.extend(ns) | ||
| self.validate_initiator(gw, ns, gw.ana_group_id) |
There was a problem hiding this comment.
In validate_initiator,
- At Initial stage, pass active gateway node object(which has to be failed over next).
- Fetch namesapces for this GW from gateway and client side.
- Validate optimized path(i.e, basically Active gateway which is passed).
- At Failover stage, collect the active gateway from failover methods.
- Use this Active GW against namespaces found in Step 2 to validate Optimized path on client is set to new active GW.
- At failback condition, collect the failed-over gateway.
- Repeat step 2.
There was a problem hiding this comment.
This is exactly the validation I have already added in the validate_initiator and ha.run methods
| empty list if the gateway is inaccessible for all devices | ||
| """ | ||
| out, _ = self.node.exec_command( | ||
| cmd="nvme list --output-format json", sudo=True |
There was a problem hiding this comment.
This is already part of nvme initiator lib, if not add this lib method.
There was a problem hiding this comment.
I saw the lib method now. Will make changes accordingly.
There was a problem hiding this comment.
This might not be required when we go with the uuid approach as discussed
| out, _ = self.node.exec_command( | ||
| cmd="nvme list --output-format json", sudo=True | ||
| ) | ||
| nvme_devices = json.loads(out).get("Devices", []) |
There was a problem hiding this comment.
At client side,
- Need to identify the right devices against the namespace passed from GW side(probably using lsblk -o name,wwn).
- Get the topology for that device(i.e get the path status for each GW path).
- Validate against the GW which has optimized path.
There was a problem hiding this comment.
nvme list-subsys gives the namespace id and subsystem name, isn't this enough proof to get the required namespaces?
| subsystems = json.loads(out)[0].get("Subsystems") | ||
| for subsys in subsystems: | ||
| for path in subsys.get("Paths"): | ||
| if gw_ip in path.get("Address") and path.get("State") == "live" and path.get("ANAState") == "optimized": |
There was a problem hiding this comment.
create separate method to list out optimized and inaccessible paths for a device.
{"device_wwn": {"optimized": []}, {"inaccessible": []}. use it for validation of paths
There was a problem hiding this comment.
will do this change as well.
Adding initiator side validation for NVMe HA to verify optimised and inaccessible paths.
Logic written:
Logs - http://magna002.ceph.redhat.com/cephci-jenkins/cephci-run-LE1K03/