An operator can create snapshots of pools

it's valuable to provide volume's snapshot management, some solution to generate snapshots, and save them to NFS. at some time, they might be restored on demand.
Note: this feature doesn't include VM snapshot.

as VSM is based openstack horizon, is it possible to embed VSM into openstack dashboard as a project?

Is it possible to add a bug to improve the performance of the sql queries.
suggested improvements:
simplification and optimization of multi layer queries for join and timestamp operations
perhaps use indexing for first timestamp range and then use primary key for each dataset
?

known adaptability issues:
i) RHEL7 uses mariadb so should be 'systemctl start mariadb''
ii) Generally the ifconfig command does not exist on RHEL7. Use the 'ip' command instead. (You can install ifconfig but its use is discouraged)

I defined the the ceph parameter "osd_max_backfills 2" in the cluster.manifest. I restarted the VSM controller on the Management node and VSM agents on the storage nodes. Afterward, i created a cluster through the VSM GUI and found the osd_max_backfills parameter was not included in ceph.conf. Consequently, the osd_max_backfills parameter had the default value of 10

[root@storage1 ~]# ceph --admin-daemon /var/run/ceph/ceph-osd.3.asok config show |egrep backfil
"osd_max_backfills": "10",
"osd_backfill_full_ratio": "0.85",
"osd_backfill_retry_interval": "10",
"osd_backfill_scan_min": "64",
"osd_backfill_scan_max": "512",
"osd_kill_backfill_at": "0",
"osd_debug_skip_full_check_in_backfill_reservation": "false",
"osd_debug_reject_backfill_probability": "0",

as indicated by the banner.

Automated migration of data from one EC pool to another EC pool. Rationale behind this is – you start with one EC scheme and to switch over to different EC sheme. Ceph doesn’t support in-band scheme change. You need to export/import.

in current feature set, some operations are dangerous, shall we apply an access control mechnism to distinguish different roles?

Generally, there will be 3 kind of roles:
administrator (deploy/build cluster...)
operator (operation and maintain cluster)
demo (read only)

the root cause is we added filesystem mount options in cluster manifest, if no the section, the cluster_manifest will fail.

An operator of field service engineer can inspect VSM logs to determine the timing and sequence of operations and errors reported by VSM. The messages include a message number. The messages are documented in English to facilitate translation.

As vsm controller is one key component to manage ceph cluster, in production environment, one requirement is to make the controller highly available.

some user requests to add RBD level management.

so far user can set file system in cluster.manifest, but can't set related mount options.

An operator can inspect/analyze up to 1 year of Ceph operational data with timestamps. (for what purpose?) (What data should be collected?)

The releases here would be easier to install on ubuntu if they were pushed to a PPA. I can set that up but it would be better if the project had an official PPA associated with it.

An operator can observe trend data for capacity utilization and performance metrics (public and cluster) for 90 minute, 90 hour, and 90 day intervals

it's expected to see below configuration management items:

1. manifest configuration (cluster.manifest/server.manifest)
2. ceph.conf configuration
3. vsm configuration
   Configuration management may or may not be a part of VSM UI.

During Ceph cluster create VSM can log the Ceph configuartion settings. Spit out the Ceph commands being issues while cluster is being setup

we noticed this issue as we were debugging our code on wip-53 dev branch. I put a comment to the vsm-discuss list asking if VSM dev team will be continuing this approach since modern ceph does not require these sections anymore.
I also asked in that thread: Does VSM read the /etc/ceph/ceph.conf file (ever)? If not, then we can contrive the missing sections during processing of our import script and simply concatenate the ceph.conf file with the dynamically discovered missing sections, using that composite as the ceph.conf data we pass to the VSM api.

Support an unattended demo mode - the operator can access all VSM pages and dialogs, but all VSM operations that could modify the cluster are disabled.

An operator can configure email alerts for a) OSDs over near full threshold, b) storage groups over near full threshold, c) servers that have become unavailable, d) OSDs that have been placed in autoout, c) too many incidents of long latency or client timeout

Seems there are two files that can be used:

virtual-storage-manager/source/vsm/bin/cluster_manifest
virtual-storage-manager/source/vsm-deploy/cluster_manifest

None is at the $PATH so whenever cluster_manifest is called, it is giving errors.

Which is the correct file to be used during install.sh and vsm-installer?

We need to add a background service (periodic task) that updates the VSM database to reflect cluster topology changes made by command-line tools and other ceph management tools.
As it stands, the only way VSM is able to deal properly with such external Ceph configuration changes is by resetting and reimporting the entire ceph cluster after making such changes.

An operator can configure SNMP alerts for a) OSDs over near full threshold, b) storage groups over near full threshold, c) servers that have become unavailable, d)OSDs that have been placed in autoout, c) too many incidents of long latency or client timeout

The operator cannot start certain operations if the cluster is temporarily in a fault condition. (need to be more specific about what fault conditions - for instance, want to be able to add a server is Ceph is reporting OSD full)

Allows pools to be deleted. This is dangerous in a production environment but useful in a POC environment. Perhaps this feature could be controlled via CLI so the Feature could be disabled when the cluster is in production.

we have MDS status check now, and user also hopes to have MDS management capabilities.

"VSM should support two operator roles:
Admin: Can perform all VSM operations
Operator: Can perform routine maintenance - restart OSDs, remove OSDs, restore OSDs"

Grant specific roles to each operator - for example create an operator with permission to create new storage groups and add servers (less likely to cause data loss or disrupt service) but without permission to remove servers (more likely to disrupt service or cause data loss)

VSM needs to work with current ICE versions

VSM doesn’t configure ceph client while building ceph cluster i.e we would have to configure the clients manually with admin/gatherkeys. It’s not bad but I think it would be great to have an option for configuring client as well with VSM.

occasionally, we see weird behaviors if some vsm processes stop, it's meaningful to show all vsm related processes status on UI to help quick fix.

the feature will add on dashboard to show all vsm critical service status, like vsm daemon, mariadb, keystone, rabbitmq, diamond...

An operator can attach Ceph pools to OpenStack Swift. TBD - Publishes pools to a configured RADOS gateway?

An administrator running a POC can remove pools from OpenStack Cinder that were previously added to Cinder via VSM.

An operator can create and configure RADOS gateway clients, and attach pools to VSM-created RADOS Gateway clients. (S3)

This feature request came from the following vsm-discuss thread:
http://vsm-discuss.33411.n7.nabble.com/VSM-horizon-framework-version-update-td128.html
Version 2.0 currently has been updated to Horizon "Juno", for features that were already in the 1.0 code base when it was on Havana, however, new post-Havana features have not been added - such as pluggable settings:
http://docs.openstack.org/developer/horizon/topics/settings.html#pluggable-settings
Pluggable settings would make extending VSM much simpler for third-parties.

if there are hundreds of rbd volumes, the expandable table will make the ui awful

this ensures that an operator cannot start a new operation while another incompatible operation is under way.

The operator cannot start operations until potentially conflicting operation already in flight are completed successfully and failed gracefully. In one implementation, Ceph executes operations fro a queue.

This refers specifically to providing more detailed problem solving information for getting VSM set up for the first time.
VSM has insufficient error information available when things go wrong. This starts with little explanation on what the various logs are for and what the messages in them show - but also extends to insufficient detailed error messages and codes available.

Display results of ceph PG DUMP STUCK on PG Status Page

An operator can identify RBD clients that are hot (busy). An operator can identify RADOS Gateway clients that tare hot (busy).

"Would be useful to have a client status page that showed the performance (iOS, mb/s) for each RBD.
Since there can be a lot RBDs, this page would require pagination (e.g. 100 RBDs per page) and sort (e.g. sort by IOPS or sort by mb/s)."