[Merged] openstack/cinder master: NetApp SolidFire: Fix RecursionError accessing undefined attributes https://review.opendev.org/c/openstack/cinder/+/802955
#┊・openstack
minor mason
Hi everyone, Just a reminder that next week is Cinder's second midcycle for the Zed release (10th August, 1400-1600 UTC) , and members of the NetApp OpenStack team will be attending as usual.
https://lists.openstack.org/pipermail/openstack-discuss/2022-August/029786.html
Topics can be included in meeting's etherpad: https://etherpad.opendev.org/p/cinder-zed-midcycles
quaint warren
Hi any Openstack experts here? There is a thread about not getting expected performance on an ASA800, and in normal circumstances we would check the iSCSI mulitpath.conf/multipath -ll. I guess Openstack manages that? Can someone help explain how MPIO config works?
We are seeing traffic getting split across cluster interconnect.
Edit: This is an internal link, but I want to post it here for anyone to review internally. https://latx.netapp.com/viewperf/id/463370
minor mason
Hi @quaint warren , sorry for the late response
yes configuration is done in multipath.conf
I'm just checking if the cinder type should include a flag to enable multipath
ok, so first thing is to have the multipath support enabled in nova, configuration option is 'volume_use_multipath'
https://docs.openstack.org/nova/yoga/configuration/config.html
there is no need to make changes to the cinder type
also if you need to customize default multipath configuration you need to create /etc/mulitpath.conf and add your settings in it
minor mason
No further configuration is necessary.
When attaching volumes to instances, Cinder will pass to Nova the list of targets, 'volume_use_multipath' option ensures that Nova will login to the targets in parallel
visual grail
We have a customer using the Openstack cinder driver https://github.com/openstack/cinder. Reviewing the docs and the audit logs, it is still using API\ZAPI. Is this going to be updated to support REST?
solid dust
Hey, @visual grail
We expect to have around 90% of all Cinder ZAPI calls moved to REST for the Zed release (October/2022). REST will be marked as an experimental feature and ZAPI will remain the default and recommended method for now.
In Antelope release (April/2023) all ONTAP drivers (Manila and Cinder) will have full support to REST and it'll become the default communication method with ONTAP.
visual grail
@solid dust Thank you very much.
worthy turtle
We are happy to announce that the ZAPI to REST transition of our Cinder drivers API calls has been merged!
The REST calls for Cinder are available in the Openstack Zed release. We still keep ZAPI calls as default.
- [Merged] NetApp ONTAP: Add REST Client for ONTAP https://review.opendev.org/c/openstack/cinder/+/836027
- [Merged] NetApp ONTAP: Add core functions on REST client https://review.opendev.org/c/openstack/cinder/+/852208
- [Merged] NetApp ONTAP: Add volume replication functions on REST client https://review.opendev.org/c/openstack/cinder/+/852209
- [Merged] NetApp ONTAP: Add volume migration functions on REST client https://review.opendev.org/c/openstack/cinder/+/852210
- [Merged] NetApp ONTAP: Add revert to snapshot functions on REST client https://review.opendev.org/c/openstack/cinder/+/852211
spiral lantern
Hi everyone, we are having an issue with max_oversubscription_ratio = 1.0 not being honored. Just checking to see if anyone is facing similar issues where its provisioning/allocating capacity more than the volume size in cinder even though oversubscription is set. We have a 10TB volume and we are seeing provisioning crossing 10TB on most of our volumes, wondering if this is a bug or expected behavior. Openstack version is RHOSP 13.
storm flume
Hi everyone, we are having an issue with max_oversubscription_ratio = 1.0 not be...
Hi @spiral lantern, sorry for the late response, I received a question related to max_oversubscription_ratio by e-mail (not sure if it's from the same person) and I'm investigating... As soon as I have an answer I'll let you know
storm flume
@spiral lantern I did a few tests in my environment and I was not able to reproduce this behavior. So, I have some questions to you:
• Are you able to create a new volume in this pool?
• Which filter class names are being used? (
- Example: scheduler_default_filters = AvailabilityZoneFilter,CapacityFilter,CapabilitiesFilter
• Which weigher class names are being for weighing hosts?
- Example: scheduler_default_weighers = CapacityWeigher
brazen root
Hello all, can trident engage File Analytics on a volume ... and how can you tell if it did?