Bug 1547819

Here are the few more inputs from Dennis Keefe

1. slabsize can be set to 32GB if the physical storage space is few TBs.
Otherwise, we can go ignore this param in the gdeploy config file.

<input_from_DennisKeefe>

In most cases using a small slab size is done when creating smaller VDO volumes and or physical space
is limited.  If you are going to have 20TB of storage then set the slab size to 32GB.  Even if the physical 
storage for a single VDO volume is a few TB, set the slabsize to 32GB. 

</input_from_DennisKeefe>

2. writepolicy should be set to 'auto'

<input_from_DennisKeefe>

The latest version of VDO has an auto option, which will select the correct
mode based on the storage below VDO.  Only set VDO to sync mode if the 
underlying storage does not have a write back cache.  My recommendation would
be to used the new auto default. 

</input_from_DennisKeefe>

3. readcache is enabled, with cachesize set to 20MB

<input_from_DennisKeefe>

My recommendation is to:
* Enable the read cache
* Set the read cache to 20M - This will help when reading blocks for faster read verify 
of duplicates and compressed blocks.

</input_from_DennisKeefe>

4. block map cache size should be initially set to 128MB and increased when required.

<input_from_DennisKeefe>

* Block map cache default size is 128MB.  This can be adjusted later if it is needed. 
I'll follow up with how to identify whether more block map cache is needed.

</input_from_DennisKeefe>

Based on the inputs about the gdeploy configuration will look like 
if the physical storage is few TBs ( say 2TB )

[vdo]
names=vdovol_sdb
devices=sdb
logicalsize=<computed as 10 x physical size of disk sdb>
# slabsize is set to 32GB for few TBs too
slabsize=32GB
blockmapcachesize=128MB
readcache=enabled
readcachesize=20MB
emulate512=enabled
writepolicy=auto
ignore_vdo_errors=no

(In reply to SATHEESARAN from comment #4)
> 
> [vdo]
> names=vdovol_sdb
> devices=sdb
> logicalsize=<computed as 10 x physical size of disk sdb>
> # slabsize is set to 32GB for few TBs too
> slabsize=32GB
> blockmapcachesize=128MB
> readcache=enabled
> readcachesize=20MB
> emulate512=enabled
> writepolicy=auto
> ignore_vdo_errors=no

There was small correction to the size arguments
Corrected conf file is below:

[vdo]
action=create
names=vdovol_sdb
devices=sdb
logicalsize=200T
slabsize=32G
blockmapcachesize=128M
readcache=enabled
readcachesize=20M
emulate512=enabled
writepolicy=sync
ignore_vdo_errors=no

As the dependent bug is already ON_QA, moving this bug also to ON_QA

Verified the bug in the below component version's successfully.

Component versions's:
gdeploy-2.0.2-23.el7rhgs.noarch
cockpit-ovirt-dashboard-0.11.20-1.el7ev.noarch

Steps:
1. Install VDO 
2. Go to the cockpit UI and check the compression and deduplication tab.
3. Proceed to the review tab.
4. Check if the below VDO parameters are available in the gdeploy conf file generated

Output:

[vdo1:10.70.45.29]
action=create
devices=sdb
names=vdo_sdb
logicalsize=200000G
blockmapcachesize=128M
readcache=enabled
readcachesize=20M
emulate512=enabled
writepolicy=sync
ignore_vdo_errors=no
slabsize=32G

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHEA-2018:3523