Thursday, 2026-01-15

*** mhen_ is now known as mhen02:52
carloss#startmeeting manila15:00
opendevmeetMeeting started Thu Jan 15 15:00:22 2026 UTC and is due to finish in 60 minutes.  The chair is carloss. Information about MeetBot at http://wiki.debian.org/MeetBot.15:00
opendevmeetUseful Commands: #action #agreed #help #info #idea #link #topic #startvote.15:00
opendevmeetThe meeting name has been set to 'manila'15:00
gouthamro/15:00
vharihi15:00
carthacahi15:00
carlosscourtesy ping: vhari carthaca Sai gireesh Kumar_T15:00
Saio/15:01
carloss#chair gouthamr vhari15:01
opendevmeetCurrent chairs: carloss gouthamr vhari15:01
Kumar_THi15:01
Anoop_Shukla_O/15:01
gireesho/15:02
carloss#topic Announcements15:02
carlossSchedule and Deadlines15:02
carloss#link https://releases.openstack.org/gazpacho/schedule.html (Gazpacho Schedule)15:03
carlosswe're two weeks away from feature proposal freeze and new driver deadline15:03
carlossmeaning: all new Manila features must be proposed and substantially completed, with unit, functional and integration tests by the end of the R-9 week. Collaborative review sessions must be proposed at this timeline, in order to speed up the review process.15:04
carlossand15:04
carlossby the end of the R-9 week, all new backend drivers for Manila must be substantially complete, with unit tests, and passing 3rd party CI. Drivers do not have to actually merge until feature freeze.15:04
carlossalright, that's all I had for $topic. Is there any other announcement you'd like to share with the community today?15:06
carloss#topic Review Focus15:08
carlosslooking at our review focus etherpad15:09
carloss#link https://etherpad.opendev.org/p/manila-gazpacho-review-focus (Gazpacho review focus)15:09
carlossnew changes were added and I was finally able to start working on the reviews for the features being proposed15:10
carlossand started with QoS specs15:10
gouthamr a word about ending this meeting early, carloss has to drop off to an appointment in 10 minutes, and I’m feeling under the weather :/ I can hang around and discuss things if needed..15:10
carlossgouthamr++15:11
vharigouthamr++ 15:11
vharithere are no new bugs to triage this week15:12
vharipls share any concerns in this slot or via the manila channel15:12
carlossonly one more note on review focus: we can use the etherpad to look at the changes. I'll keep doing so today and in the upcoming week15:13
Anoop_Shukla_We wanted to talk about the same snapshot expectations from manila point of view. @Kumar_T - do you want to summarise the issue we are currently facing?15:14
SaiI think Kumar has already summarized it to carloss offline. 15:16
SaiThat should be fine I guess, today we have limited time :) 15:16
carlossthe context for everyone would be good though15:16
carlossbut yes, if you'd like, we can also continue this in #openstack-manila15:17
Kumar_TAt present ONTAP sync policy doesnt replication of  the snapshots taken prior to relationship creation15:18
Sai> the context for everyone would be good though15:18
SaiAck15:18
gouthamrfrom what I understand, snapshots taken prior to replica creation aren’t replicated when using synchronous replicas15:18
Kumar_Tfor sync policies15:18
Kumar_TWe will be raising a dependency on ontap to provide option to copy the local snapshot at the time of snapmirror init or bulk snapshot copy api15:18
Kumar_Tfor now we will be documenting as we are aligning with ontap behavior and provide workaround of manually copy using ontap api if any users wants it to do15:19
Kumar_T8:36:52 PM 15:19
Kumar_Tonce ontap gives the option we will support that usecase15:19
gouthamrokay, doesn’t the driver have a mechanism to report individual snapshot status? if so it can set it to “error” and have a user message indicating why15:19
Kumar_TYes15:19
Kumar_Twe provide steps to copy manually15:20
Kumar_Tonce it is done it should be available and in-sync in next discovery15:20
gouthamroh if there’s a workaround, wouldn’t the driver just want to do it?15:20
gouthamrcarloss: I can take it from here, Ty!15:20
carlossgouthamr: thank you very much15:21
Anoop_Shukla_Why is there an expectation from manila that all snapshots should be copied? Is that how all other storage policies for snapmirror work as well?15:21
Kumar_Tthe existing transfers api comes with limitation of single copy and no parallel calls15:21
Anoop_Shukla_I mean for other vendor drivers.15:21
gouthamrAnoop_Shukla_: yes that’s how it was designed; but, with flexibility iirc to set an “error” state where a snapshot couldn’t be replicated15:23
carlossgouthamr++ Anoop_Shukla_++ Kumar_T++ vhari++ Sai++ thanks folks15:24
gouthamrTy carloss15:24
Anoop_Shukla_Is the expectation that a backup is unusable if it cannot be replicated?15:24
Anoop_Shukla_Thanks Carlos take care.15:24
Anoop_Shukla_Because these are two different use cases. One for local protection another one for remote protection.15:25
Kumar_TTy Carlos15:25
SaiTake care and thanks carloss15:25
gouthamryes, imagine the primary site is down, you promote to use the secondary site and expect to still use your backup15:26
vharicarloss++15:26
Anoop_Shukla_If we error out - a snapshot taken prior to replication cannot be used to revert share content.15:26
Anoop_Shukla_Agreed..15:26
gouthamrby backup here I mean a snapshot15:26
gouthamrA “point in time backup copy” as it is referred to sometimes15:27
Anoop_Shukla_Yep..customers generally keep golden copies of backups..which serve as goto snapshots in worst case.15:27
gouthamrthere is a natural queuing mechanism implemented in the creation of replicated snapshots15:28
gouthamrthe manager polls the driver repeatedly to check whether the snapshot has made it15:28
gouthamrthe ONTAP driver could rely on this and wait for a replica to go in-sync before attempting to replicate the snapshot15:29
Anoop_Shukla_So in case currently we do not have a way to efficiently copy older snapshots with sync snapmirror..what should be our way forward? Should we still do it via snapmirror transfers api one by one..which can be painstakingly slow sometimes.15:29
gouthamryes, you’re suggesting that as a workaround, aren’t you15:30
Anoop_Shukla_Yep..since we cannot do it during initialize..we will need to perform it in looped call..15:30
gireeshthis should be one time operation, do we need looping call for this 15:32
gouthamryeah, consider that.. there may be some kinks to iron out in the share manager too wrt the status updates15:32
Anoop_Shukla_We will open a requirement to the platform team..but till that's not there..we may need to use this workaround15:32
gouthamrthe looping call runs from the share manager15:32
Anoop_Shukla_Right15:32
Anoop_Shukla_Kumar any other views on this?15:33
Kumar_TI dont have any, we need to go with manual steps for time being until either ontap support for sync or some other means platform you are referring.15:34
gouthamrby manual steps, you mean Apis explicitly executed by the driver after waiting15:35
Anoop_Shukla_Manual you mean we will do transfers within the driver code right?15:35
Kumar_Tnot in code as parallel calls and bulk transfers are not allowed by SM on same target.15:36
Kumar_Tusing ontap cli or api.15:36
Kumar_Tfor those precreated snapshots15:36
gouthamrI’ll let you check how snapshots are replicated in the share manager before making that call15:38
gouthamralright, any other topics for today?15:40
gouthamrlooks like nothing else; if you have any please post on #openstack-manila15:42
gouthamrthank you for participating, apologies for the shorter meeting today15:42
SaiThank you, take care gouthamr !! 15:42
Kumar_TThank you all.15:42
gouthamr#endmeeting15:42
opendevmeetMeeting ended Thu Jan 15 15:42:50 2026 UTC.  Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)15:42
opendevmeetMinutes:        https://meetings.opendev.org/meetings/manila/2026/manila.2026-01-15-15.00.html15:42
opendevmeetMinutes (text): https://meetings.opendev.org/meetings/manila/2026/manila.2026-01-15-15.00.txt15:42
opendevmeetLog:            https://meetings.opendev.org/meetings/manila/2026/manila.2026-01-15-15.00.log.html15:42
gouthamrthank you Sai15:42
*** gmaan is now known as gmaan_afk16:33
*** gmaan_afk is now known as gmaan17:46

Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!