FC Storage challenges

J

jone

New Member
Jul 11, 2024
7
0
1
I am on a path to migrate 15x ESXi servers (With vcenter / Distributed networking / cluster) and 200TB FC SSD SAN, with Proxmox virtualization. I am facing lots of challenges with the lack of a "good filesystem" like VMFS.

I have my first 2 PVE servers up running with Multipath, and LVM in shared mode. The performance is really good, and migration of Virtual machines between the nodes is working really well.
Some challenges that I am trying to get some ideas on.
1. How do I easiest make sure that we maintain identical Multipath config / Disk naming across all servers? If I make a change / provision a new LUN, I will have to modify the Multipath files on EVERY server.
2. How can I create a shared LUN and use that as a source folder for ISO / Installation media and have it available for all nodes? Looks like I cannot upload ISO files to my LVM drive, and not create folder on LVM.
3. How can I easiest expand a LUN without downtime?

These were all tasks that were quick and seamless on ESXi.
 
jone said:
1. How do I easiest make sure that we maintain identical Multipath config / Disk naming across all servers? If I make a change / provision a new LUN, I will have to modify the Multipath files on EVERY server.
Click to expand...
Automation is the worst enemy of human error. Implement Ansible and Version Control if you want to to avoid copy/pasting.
jone said:
How can I create a shared LUN and use that as a source folder for ISO / Installation media and have it available for all nodes? Looks like I cannot upload ISO files to my LVM drive, and not create folder on LVM.
Click to expand...
You either need to deploy a Cluster Aware Filesystem (ie OCFS2), or pass the LUN to a VM that runs NFS server
jone said:
How can I easiest expand a LUN without downtime?
Click to expand...
Perhaps you can expand on what you have done so far that caused downtime?
FC LUN expansion happens on your storage side, the host should automagically see larger raw LUN size. You will then need to expand all involved layers (PV , VG, LV pool). In general all of this happens without downtime.

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
 
Last edited:
1. As luns have unique id's you can put all luns in the multipath config at once and even distribute to all servers the same.
2. Create a filesystem on a lun after which you can create a folder for your iso's and provide it in the datacenter for all pve's.
3. Extend the lun in your storage then use lvextend. If that lvm volume has a filesystem resize that also.
 
waltar said:
1. As luns have unique id's you can put all luns in the multipath config at once and even distribute to all servers the same.
2. Create a filesystem on a lun after which you can create a folder for your iso's and provide it in the datacenter for all pve's.
3. Extend the lun in your storage then use lvextend. If that lvm volume has a filesystem resize that also.
Click to expand...
For 2, Are there any File systems that can be used on a FC Storage, and supports shared (That anyone would advice on using)?
I am looking at either install an ISCSI system that I have in addition to FC, OR run FreeNAS as ISCSI for sharing of files.
 
bbgeek17 said:
Automation is the worst enemy of human error. Implement Ansible and Version Control if you want to to avoid copy/pasting.

You either need to deploy a Cluster Aware Filesystem (ie OCFS2), or pass the LUN to a VM that runs NFS server

Perhaps you can expand on what you have done so far that caused downtime?
FC LUN expansion happens on your storage side, the host should automagically see larger raw LUN size. You will then need to expand all involved layers (PV , VG, LV pool). In general all of this happens without downtime.

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
Click to expand...
Been doing some more testing.
Expanded my test LUN from 100G to 200G on my storage system. This is shared to 2 different systems, and using multipath.

root@SVR-DC-PX1:~# multipath -ll
3PARPV1 (360002ac0000000000000032c0001fdcd) dm-5 3PARdata,VV
size=30T features='1 queue_if_no_path' hwhandler='1 alua' wp=rw
`-+- policy='round-robin 0' prio=50 status=active
|- 1:0:1:0 sde 8:64 active ready running
|- 2:0:0:0 sdg 8:96 active ready running
|- 1:0:0:0 sdc 8:32 active ready running
`- 2:0:1:0 sdi 8:128 active ready running
3PARPV2 (360002ac000000000000003370001fdcd) dm-6 3PARdata,VV
size=100G features='1 queue_if_no_path' hwhandler='1 alua' wp=rw
`-+- policy='round-robin 0' prio=50 status=active
|- 1:0:1:1 sdf 8:80 active ready running
|- 2:0:0:1 sdh 8:112 active ready running
|- 1:0:0:1 sdd 8:48 active ready running
`- 2:0:1:1 sdj 8:144 active ready running


Rescanned devices, no change. Restarted Multipath service, and the new size shows up.
root@SVR-DC-PX1:~# multipath -ll
3PARPV1 (360002ac0000000000000032c0001fdcd) dm-5 3PARdata,VV
size=30T features='1 queue_if_no_path' hwhandler='1 alua' wp=rw
`-+- policy='round-robin 0' prio=50 status=active
|- 1:0:1:0 sde 8:64 active ready running
|- 2:0:0:0 sdg 8:96 active ready running
|- 1:0:0:0 sdc 8:32 active ready running
`- 2:0:1:0 sdi 8:128 active ready running
3PARPV2 (360002ac000000000000003370001fdcd) dm-6 3PARdata,VV
size=200G features='1 queue_if_no_path' hwhandler='1 alua' wp=rw
`-+- policy='round-robin 0' prio=50 status=active
|- 1:0:1:1 sdf 8:80 active ready running
|- 2:0:0:1 sdh 8:112 active ready running
|- 1:0:0:1 sdd 8:48 active ready running
`- 2:0:1:1 sdj 8:144 active ready running
Trying to Resize the Multipath device, but it fails.
root@SVR-DC-PX1:~# multipathd -k"resize map /dev/mapper/3PARV2"
fail

My previous attempt was solved by a reboot on all shared nodes.
Hoping to find a flow that is working seamless.

Thanks
 
jone said:
Hoping to find a flow that is working seamless.
Click to expand...
There are many variables involved. Things like "holders", IO load, etc.

You can try this procedure https://docs.redhat.com/en/document...ath/online_device_resize#online_device_resize (second google result for me "multipath new size" ).

You may also want to consult HPE/3PAR support/site for best approach. Its not PVE specific, there are bugs in software:

https://bugzilla.redhat.com/show_bug.cgi?id=352421#c16
https://dm-devel.redhat.narkive.com...ath-maps-reload-ioctl-failed-invalid-argument

Good luck

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
 
jone said:
1. How do I easiest make sure that we maintain identical Multipath config / Disk naming across all servers? If I make a change / provision a new LUN, I will have to modify the Multipath files on EVERY server.
Click to expand...
Why do you require the multipath file to be customized? I have implemented some clusters at customers and I only mask the local disks and allow all disks with the correct vendor string.
jone said:
2. How can I create a shared LUN and use that as a source folder for ISO / Installation media and have it available for all nodes? Looks like I cannot upload ISO files to my LVM drive, and not create folder on LVM.
Click to expand...
You can use OCFS, but I prefer to use NFS for ISO files.
jone said:
3. How can I easiest expand a LUN without downtime?
Click to expand...
I have very mixed experiences. With some storages it's a bit tricky and with others it works better.
Often it is enough to run rescan-scsi-bus.shh once and then multipath -v3 once.
But I have already seen several times that the extension was only successful after all nodes had been restarted once.

I generally size the LUNs a little larger, the storages do thin provisioning anyway.
I usually create a new additional LUN, which is easier and faster.

However, I also prepare these customers for a switch to Ceph when they purchase new nodes. Then life is much easier.
 
jone said:
1. How do I easiest make sure that we maintain identical Multipath config / Disk naming across all servers? If I make a change / provision a new LUN, I will have to modify the Multipath files on EVERY server.
Click to expand...
For the most part, once you set up your multipath config file, it should work the same way on all hosts using it. this isnt really a problem, usually even the default config works "well enough." mostly issues arise due to lvm and multipath race conditions at boot, which are solvable with the multipath-boot-tools package.
jone said:
2. How can I create a shared LUN and use that as a source folder for ISO / Installation media and have it available for all nodes? Looks like I cannot upload ISO files to my LVM drive, and not create folder on LVM.
Click to expand...
create a nas bridge vm serving out nfs. normally I dont bother with this; your nas can and should be external to your main cluster- I usually either use the backup device, or a separate device alltogether.

jone said:
How can I easiest expand a LUN without downtime?
Click to expand...
afaik resizing a lun wont cause downtime anyway, and a simple Echo "- - -" > /sys/class/scsi_host/hostX/scan should cause the host to see the new capacity.

What you arent asking is just as important. There isnt a simple way to expose snapshot functionality for scsi based storage, if you're going to use your storage for snapshots, it will not be exposed to the host os, nor will it have exposed filesystem quiescence- meaning you'll need to manually orchestrate this both for taking snapshots and for restoration of any kind.
 
alexskysilk said:
What you arent asking is just as important. There isnt a simple way to expose snapshot functionality for scsi based storage, if you're going to use your storage for snapshots, it will not be exposed to the host os, nor will it have exposed filesystem quiescence- meaning you'll need to manually orchestrate this both for taking snapshots and for restoration of any kind.
Click to expand...
Yes, snapshots do not work on LVM, but you can still make snapshot backups, which is also very fast with the PBS. This is usually sufficient as a replacement for VM snapshots.

But why shouldn't quiescing work? If the qemu guest agent is installed, it works normally.
 
Yes, snapshots in storage, but the problem exists in many environments and is often only solved via agents in the hosts.
 
Falk R. said:
But many customers don't use the functions at all
Click to expand...
I LOVE customers who pay for stuff they dont use! my favorite!

snapshot integration is a pretty big deal in any vsphere infrastructure I support. If op doesnt want/need/use it, then its not an issue. the good news is that implementing it (at least in a basic form) is a simple script that issues a vm freeze, calls snapshot create api, and vm thaw, so the orchestration isnt complicated- but it does not come in the box, either from proxmox or the storage vendor.
 
alexskysilk said:
afaik resizing a lun wont cause downtime anyway, and a simple Echo "- - -" > /sys/class/scsi_host/hostX/scan should cause the host to see the new capacity.
Click to expand...
Worst case is to rescan on every level:
  • rescan all disks (echo 1 > /sys/block/sda/device/rescan) for alle devices in your multipath or on your host - or use rescan-scsi-bun.sh from scsitools
  • reload multipath (multipath -r or systemctl reload multipathd)
  • pvresize the affected physical volumes
 
LnxBil said:
Worst case is to rescan on every level:
  • rescan all disks (echo 1 > /sys/block/sda/device/rescan) for alle devices in your multipath or on your host - or use rescan-scsi-bun.sh from scsitools
  • reload multipath (multipath -r or systemctl reload multipathd)
  • pvresize the affected physical volumes
Click to expand...
I spent some time over the weekend and tested. My test LUN grew from 100G to 350G, but like it have been mentioned, its thin provisioned on my 3PAR. I have to rescan, reload, and pvresize. Now that I have that sorted out, everything gets easier. Now I need to change the Multipath file to wildcard. This is a long way to go coming from vcenter / esxi, but its free :D
 
Falk R. said:
Why do you require the multipath file to be customized? I have implemented some clusters at customers and I only mask the local disks and allow all disks with the correct vendor string.

You can use OCFS, but I prefer to use NFS for ISO files.

I have very mixed experiences. With some storages it's a bit tricky and with others it works better.
Often it is enough to run rescan-scsi-bus.shh once and then multipath -v3 once.
But I have already seen several times that the extension was only successful after all nodes had been restarted once.

I generally size the LUNs a little larger, the storages do thin provisioning anyway.
I usually create a new additional LUN, which is easier and faster.

However, I also prepare these customers for a switch to Ceph when they purchase new nodes. Then life is much easier.
Click to expand...
With Multipath, my experience has been to use multipath config file with the wwid and an alias. This requires each LUN to have a static config on every node.

Like this:
multipaths {
multipath {
wwid "360002ac0000000000000032c0001fdcd"
alias 3PARData1
}

If I am using wildcard config, will I be able to add new LUN's without edit the multipath file?
Anywone that have an example file to show me / educate me?

Thank you
 
You must log in or register to reply here.

About

The Proxmox community has been around for many years and offers help and support for Proxmox VE, Proxmox Backup Server, and Proxmox Mail Gateway.
We think our community is one of the best thanks to people like you!

Quick Navigation

Home Get Subscription Wiki Downloads Proxmox Customer Portal About

Get your subscription!

The Proxmox team works very hard to make sure you are running the best software and getting stable updates and security enhancements, as well as quick enterprise support. Tens of thousands of happy customers have a Proxmox subscription. Get yours easily in our online shop.

Buy now!
Community platform by XenForo® © 2010-2024 XenForo Ltd.
Top Bottom
Back