Bug 1667031

Summary:	heketi is not adding new node in existing volume's "backup-volfile-servers"
Product:	[Red Hat Storage] Red Hat Gluster Storage	Reporter:	Nitin Goyal <nigoyal>
Component:	heketi	Assignee:	John Mulligan <jmulligan>
Status:	CLOSED DUPLICATE	QA Contact:	Aditya Ramteke <aramteke>
Severity:	high	Docs Contact:
Priority:	high
Version:	ocs-3.11	CC:	aramteke, hchiramm, jmulligan, knarra, kramdoss, madam, ndevos, nigoyal, rhs-bugs, rtalur, sankarshan, storage-qa-internal
Target Milestone:	---
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	If docs needed, set a value
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2019-07-25 14:15:53 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Nitin Goyal 2019-01-17 09:52:06 UTC

Description of problem:
when we are adding new node in heketi it is not adding that node in existing volumes option "Mount Options: backup-volfile-servers".

Version-Release number of selected component (if applicable):
8.0.0-7.el7rhgs.x86_64

How reproducible:


Steps to Reproduce:
1. add new node into heketi
2. check the volume info of any existing volume.
3. 

Actual results:
It is not adding node in backup-volfile-servers

Expected results:
It should add node in backup-volfile-servers

Additional info:

Comment 4 Niels de Vos 2019-01-17 15:13:27 UTC

I do not think there is a good reason to add the newly added host to the backup-volfile-servers for existing volumes. The newly added servers do not provide additional redundancy or higher availability, as those servers do not contain bricks with data of those volumes. Without the majority of the 'old' backup-volfile-servers online, the volume will not have sufficient bricks available to be used. (In that case, mounting should ideally fail so that kubernetes re-spins the pod later.)

What is the reason you think the backup-volfile-servers should include all the storage servers?

Comment 5 Nitin Goyal 2019-01-18 05:35:01 UTC

Hi Niels,

I have raised this bug because there is inconsistencies because of this issue in heketidb, If we don't want to add new host in backup-volfile-servers then we should handle this efficiently in "heketi-cli db check".

Comment 6 John Mulligan 2019-01-18 14:31:20 UTC

(In reply to Nitin Goyal from comment #5)
> Hi Niels,
> 
> I have raised this bug because there is inconsistencies because of this
> issue in heketidb, If we don't want to add new host in
> backup-volfile-servers then we should handle this efficiently in "heketi-cli
> db check".

First, unless you are statically provisioning the volumes I don't think the "backup-volfile-servers" is in play in current deployments. I'd like Talur to confirm that my statement is true.

Next, if we are somehow using the parameter I can understand why one would want to keep the list up to date with nodes in the cluster. 

Finally, I don't think "db check" would be involved here as it is purely an internal consistency check and must only relate to the contents of the db as heketi is implemented, not as we might desire the content to be. In other words, we check for the linkage between a volume and a brick but not that device size in the db matches the device size on the node. There are other components for that kind of thing.

Comment 7 Yaniv Kaul 2019-04-01 06:54:24 UTC

(In reply to John Mulligan from comment #6)
> (In reply to Nitin Goyal from comment #5)
> > Hi Niels,
> > 
> > I have raised this bug because there is inconsistencies because of this
> > issue in heketidb, If we don't want to add new host in
> > backup-volfile-servers then we should handle this efficiently in "heketi-cli
> > db check".
> 
> First, unless you are statically provisioning the volumes I don't think the
> "backup-volfile-servers" is in play in current deployments. I'd like Talur
> to confirm that my statement is true.
> 
> Next, if we are somehow using the parameter I can understand why one would
> want to keep the list up to date with nodes in the cluster. 
> 
> Finally, I don't think "db check" would be involved here as it is purely an
> internal consistency check and must only relate to the contents of the db as
> heketi is implemented, not as we might desire the content to be. In other
> words, we check for the linkage between a volume and a brick but not that
> device size in the db matches the device size on the node. There are other
> components for that kind of thing.

Status?

Comment 8 John Mulligan 2019-04-01 14:13:32 UTC

(In reply to Yaniv Kaul from comment #7)
> (In reply to John Mulligan from comment #6)
> > (In reply to Nitin Goyal from comment #5)
> > > Hi Niels,
> > > 
> > > I have raised this bug because there is inconsistencies because of this
> > > issue in heketidb, If we don't want to add new host in
> > > backup-volfile-servers then we should handle this efficiently in "heketi-cli
> > > db check".
> > 
> > First, unless you are statically provisioning the volumes I don't think the
> > "backup-volfile-servers" is in play in current deployments. I'd like Talur
> > to confirm that my statement is true.
> > 
> > Next, if we are somehow using the parameter I can understand why one would
> > want to keep the list up to date with nodes in the cluster. 
> > 
> > Finally, I don't think "db check" would be involved here as it is purely an
> > internal consistency check and must only relate to the contents of the db as
> > heketi is implemented, not as we might desire the content to be. In other
> > words, we check for the linkage between a volume and a brick but not that
> > device size in the db matches the device size on the node. There are other
> > components for that kind of thing.
> 
> Status?

Still need to confirm what "backup-volfile-servers" is used for. Forgot to needinfo on Talur last time. 

Once we have a better idea of the impact of a change in this area we can decide if it needs to be done and when.

Comment 9 Yaniv Kaul 2019-05-02 09:41:45 UTC

(In reply to John Mulligan from comment #8)
> (In reply to Yaniv Kaul from comment #7)
> > (In reply to John Mulligan from comment #6)
> > > (In reply to Nitin Goyal from comment #5)
> > > > Hi Niels,
> > > > 
> > > > I have raised this bug because there is inconsistencies because of this
> > > > issue in heketidb, If we don't want to add new host in
> > > > backup-volfile-servers then we should handle this efficiently in "heketi-cli
> > > > db check".
> > > 
> > > First, unless you are statically provisioning the volumes I don't think the
> > > "backup-volfile-servers" is in play in current deployments. I'd like Talur
> > > to confirm that my statement is true.
> > > 
> > > Next, if we are somehow using the parameter I can understand why one would
> > > want to keep the list up to date with nodes in the cluster. 
> > > 
> > > Finally, I don't think "db check" would be involved here as it is purely an
> > > internal consistency check and must only relate to the contents of the db as
> > > heketi is implemented, not as we might desire the content to be. In other
> > > words, we check for the linkage between a volume and a brick but not that
> > > device size in the db matches the device size on the node. There are other
> > > components for that kind of thing.
> > 
> > Status?
> 
> Still need to confirm what "backup-volfile-servers" is used for. Forgot to
> needinfo on Talur last time. 
> 
> Once we have a better idea of the impact of a change in this area we can
> decide if it needs to be done and when.

Update?

Comment 19 John Mulligan 2019-07-25 14:15:53 UTC


*** This bug has been marked as a duplicate of bug 1660681 ***