Bug 1319579 - [Doc RFE] Enable sharding for VM store usecase
Summary: [Doc RFE] Enable sharding for VM store usecase
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: Red Hat Gluster Storage
Classification: Red Hat
Component: doc-Configuring_Red_Hat_Enterprise_Virtualization_with_RHGS
Version: rhgs-3.1
Hardware: All
OS: Linux
unspecified
high
Target Milestone: ---
: RHGS 3.1.3
Assignee: Laura Bailey
QA Contact: SATHEESARAN
URL:
Whiteboard:
Depends On:
Blocks: 1311845 1317852
TreeView+ depends on / blocked
 
Reported: 2016-03-21 05:56 UTC by SATHEESARAN
Modified: 2016-06-29 14:20 UTC (History)
8 users (show)

Fixed In Version:
Doc Type: Enhancement
Doc Text:
Clone Of:
Environment:
Last Closed: 2016-06-29 14:20:23 UTC
Target Upstream Version:


Attachments (Terms of Use)

Description SATHEESARAN 2016-03-21 05:56:49 UTC
Description of problem:
-----------------------
Sharding feature[1] will be available/supported with RHGS 3.1.3, which will be useful for VM store usecases.

This information should be recommended/must for using RHGS volume for VM store usecase.

[1] - http://blog.gluster.org/2015/12/introducing-shard-translator/

Doc
----
https://access.redhat.com/documentation/en-US/Red_Hat_Storage/3.1/html/Configuring_Red_Hat_Enterprise_Virtualization_with_Red_Hat_Gluster_Storage/chap-Hosting_Virtual_Machine_Images_on_Red_Hat_Storage_volumes.html

How reproducible:
-----------------
N/A

Additional info:
----------------
1. Shard-block-size should be as per the recommendation from perf team

Comment 1 SATHEESARAN 2016-03-21 06:04:52 UTC
Info about shard from 'volume set help'
[root@ ~]# gluster volume set help | grep shard -A4

Option: features.shard-block-size
Default Value: 4MB
Description: The size unit used to break a file into multiple chunks

Enable sharding on the volume.
# gluster volume set <vol-name> features.shard enable

There should be important note saying that - "When sharding feature is turned on, it doesn't shards the existing VM Images. The newly created VM Images file post enabling sharding will only be sharded.

Comment 10 Krutika Dhananjay 2016-05-12 05:58:03 UTC
Looks good, Laura.

Comment 12 SATHEESARAN 2016-05-12 13:34:54 UTC
Laura,

Here are the feedback about the content

1. The point about configuring shard starts with the line - 'optional' - With RHGS 3.1.3, we are recommending sharding for the volumes backing RHEV data domain
So can we remove the phrase - 'Optional: Configure sharding' as 'Configure sharding'

2. Its really a valuable thought for adding sharding related options in virt profile. Appreciate it ! Its recommended to have keys along with their namespace

<snip>
If you want to enable sharding for your entire virt group, you can add the following lines to the /var/lib/glusterd/groups/virt file.
shard=enable
shard-block-size=512MB
data-self-heal-algorithm=full
</snip>

should be

<snip>
If you want to enable sharding for your entire virt group, you can add the following lines to the /var/lib/glusterd/groups/virt file.
features.shard=enable
features.shard-block-size=512MB
cluster.data-self-heal-algorithm=full
</snip>

Rest of the content looks good to me

@Laura -  could you address the above mentioned comments ?

Comment 13 Laura Bailey 2016-05-13 01:52:30 UTC
Hi Sas,

1. Do we require sharding to be enabled in order to support this use case? If so, then yes, I agree we should remove 'Optional' entirely.

However, if sharding is a recommendation rather than a requirement (which I think is the case), perhaps we can change 'Optional' to 'Recommended' here - this indicates that sharding is not a required part of the system, but it is good to have.

I've made the change to 'Recommended' in the meantime.

Can you confirm that this is a good decision?


2. I've updated the step to include the namespaces for these options.

For consistency, I'd like to add namespaces to the list of options in groups/virt file, but some of them don't appear in our 'Configuring Volume Options' section.

Can you tell me the namespaces for these, and confirm that the namespaces I do have are correct?

performance.quick-read=off
read-ahead=off <-- should this be readdir-ahead? neither is covered in 'Configuring Volume Options'
io-cache=off <-- not covered in 'Configuring Volume Options'
stat-prefetch=off <-- not covered in 'Configuring Volume Options'
eager-lock=enable <-- not covered in 'Configuring Volume Options'
remote-dio=on <-- not covered in 'Configuring Volume Options'
cluster.quorum-type=auto
cluster.server-quorum-type=server

Comment 14 SATHEESARAN 2016-05-23 05:15:57 UTC
> Hi Sas,
> 
> 1. Do we require sharding to be enabled in order to support this use case?
> If so, then yes, I agree we should remove 'Optional' entirely.
> 
> However, if sharding is a recommendation rather than a requirement (which I
> think is the case), perhaps we can change 'Optional' to 'Recommended' here -
> this indicates that sharding is not a required part of the system, but it is
> good to have.
> 
> I've made the change to 'Recommended' in the meantime.
> 
> Can you confirm that this is a good decision?

As I understand, from RHGS 3.1.3, sharding will be the required part of RHEV-RHGS integration stack, as sharding provides better performance and also
faster self-heal. 

@Sayan, what do you think about this ?
 - Sharding should be mentioned as the requirement, isn't it ?

Comment 16 SATHEESARAN 2016-05-23 05:19:57 UTC
I am moving this bug to ASSIGNED state, as this bug requires some additional update to be done to the modified content

Comment 18 SATHEESARAN 2016-05-24 14:23:12 UTC
(In reply to SATHEESARAN from comment #16)
> I am moving this bug to ASSIGNED state, as this bug requires some additional
> update to be done to the modified content

I moved this bug to FailedQA just to accomodate the changes as mentioned in comment13. The changes are already available in the content, except for sharding mentioned as the 'recommended' practice

No other changes required

Comment 21 SATHEESARAN 2016-05-25 07:36:47 UTC
verified in the admin_doc and also in RHEV_RHGS guide with the doc_link available in comment20


Note You need to log in before you can comment on or make changes to this bug.