XCP-ng 8.3 updates announcements and testing

anthoineb

@MajorP93 Try first to fix the ping issue with jumboframe before anything. If the issue persists then:
what are the types of your SRs unable to be mounted? Is there any error in SMlog or xensouce.log?

gduperrey

Hello @MajorP93,

Several things in your message raise questions for me.

First, this SR:

1x Linstor vSAN iSCSI configured as QCOW2

What is this storage and how is it configured?
Currently, we do not support QCOW2 on Linstor/Xostor. So, is it simply an iSCSI device to which you have applied the QCOW2 format, or is it a Xostor on the XCP-ng pool?

Regarding jumbo frames, we do not use them internally and are not currently testing them. It is often recommended not to use them for the management interface, as indicated in the XCP-ng documentation: https://docs.xcp-ng.org/networking/

Non-standard MTUs (such as jumbo frames) are not supported on management interfaces. Using them can lead to serious issues, including failed pool member joins or unexpected network outages.

Is that the case here?

bleader

@majorp93 no, these updates should not change anything on the MTU behavior. From your screenshots and ip a output, I assume eth4 is the only link used, than your vlan1060 network is indeed your management network. I guess with some confidence that:

eth4 is your actual NIC
xenbr4 is the pool-wide network without vlan using eth4
xapi1 is the bridge for vlan1060
If that's right indeed on your host side, the MTUs are properly set.

As Gaël said, jumbo on management network is not officially supported, because we always end up in situation similar to yours, where something stops working for "some" reason

But as you said, when everything is setup properly, it does work.

I have test hosts at home up to date but no similar setup to yours for SR. I tried with a pool-wide and a pool-wide + vlan as management with 9000 and the ping -M do -s 8972 does work fine in both cases, so nothing I can see here.

When it fails, what do you see? Message too long? mtu=1500 ? something else?

Do you have multiple hosts in that pool? Can you try that between the hosts as well and not toward the storage systems? I would also check on ovs with ovs-vsctl list interface that each of eth4, xennbr4 and xapi1 report 9000 mtu.

I know you said it was properly setup everywhere, but I would still be tempted to think there is a setup issue somewhere.

MajorP93

@anthoineb @gduperrey @bleader
Hi!

Thank you very much for your responses.

I actually found the issue and it was not caused by the XCP-ng patches.
Appearently one of my switches had a malfunction and lost it's jumbo frames config on the ports involved...
It seems like this happened in the time frame between XCP-ng updates .
Last time I used this setup the jumbo frames / storage setup was working fine so I thought it might be related to these patches.

Anyways sorry for taking your time in this regard-

Best regards

marcoi

for the latest updates. the test pool didnt move the vms when i went to reboot. so i moved them manually. rebooted master, move vms back patched and reboot 2nd test server in the pool.
Skipped trying rolling reboot in prod and followed the same process. both pools went well without other issues.

gleh

New maintenance update candidates for XCP-ng 8.3 LTS

This batch of updates focuses on storage bug fixes and performance improvements.

What changed

Storage

blktap:
When CBT is enabled, an issue occurred where tapdisk opened cbtlog instead of the primary QCOW2 image. This previously caused commit operations (ENOTSUPP) and leaf-coalesce routines to fail, resulting in excessive storage consumption on the SR. Tapdisk now correctly bypasses the cbtlog image to allow proper coalescing.
sm: Fix a bug where the coalesce on a secondary host of a QCOW2 VDI on LVMSR could corrupt LVM metadata and would need manual intervention to restore them.
xapi:
- Migration speeds are now improved for sparse QCOW2-backed VDIs, as empty sectors are no longer transferred.
- Preserve VDI tags on migration. This is a foundational fix for kubernetes integration.

Versions

blktap: 3.55.5-9.3.xcpng8.3 -> 3.55.5-9.4.xcpng8.3
sm: 3.2.12-23.4.xcpng8.3 -> 3.2.12-23.5.xcpng8.3
xapi: 26.1.16-1.1.xcpng8.3 -> 26.1.16-1.2.xcpng8.3

Test on XCP-ng 8.3

yum clean metadata --enablerepo=xcp-ng-testing,xcp-ng-candidates
yum update --enablerepo=xcp-ng-testing,xcp-ng-candidates
reboot

The usual update rules apply: pool coordinator first, etc.

What to test

As usual, normal use and anything else you want to test.

Test window before official release of the updates

~24 hrs

We would like to thank users who shared feedback since our last call for testing: @Andrew, @JeffBerntsen, @MajorP93, @XCP-ng-JustGreat, @acebmxer, @bufanda, @flakpyro

bufanda

@gleh installed on my lab pool. Haven’t seen any issues so far.

acebmxer

@gleh

Installed...

Updated:
  blktap.x86_64 0:3.55.5-9.4.xcpng8.3                forkexecd.x86_64 0:26.1.16-1.2.xcpng8.3                
  message-switch.x86_64 0:26.1.16-1.2.xcpng8.3       qcow-stream-tool.x86_64 0:26.1.16-1.2.xcpng8.3         
  rrdd-plugins.x86_64 0:26.1.16-1.2.xcpng8.3         sm.x86_64 0:3.2.12-23.5.xcpng8.3                       
  sm-cli.x86_64 0:26.1.16-1.2.xcpng8.3               sm-fairlock.x86_64 0:3.2.12-23.5.xcpng8.3              
  squeezed.x86_64 0:26.1.16-1.2.xcpng8.3             varstored-guard.x86_64 0:26.1.16-1.2.xcpng8.3          
  vhd-tool.x86_64 0:26.1.16-1.2.xcpng8.3             wsproxy.x86_64 0:26.1.16-1.2.xcpng8.3                  
  xapi-core.x86_64 0:26.1.16-1.2.xcpng8.3            xapi-nbd.x86_64 0:26.1.16-1.2.xcpng8.3                 
  xapi-rrd2csv.x86_64 0:26.1.16-1.2.xcpng8.3         xapi-storage-script.x86_64 0:26.1.16-1.2.xcpng8.3      
  xapi-tests.x86_64 0:26.1.16-1.2.xcpng8.3           xapi-xe.x86_64 0:26.1.16-1.2.xcpng8.3                  
  xcp-networkd.x86_64 0:26.1.16-1.2.xcpng8.3         xcp-rrdd.x86_64 0:26.1.16-1.2.xcpng8.3                 
  xenopsd.x86_64 0:26.1.16-1.2.xcpng8.3              xenopsd-cli.x86_64 0:26.1.16-1.2.xcpng8.3              
  xenopsd-xc.x86_64 0:26.1.16-1.2.xcpng8.3

MajorP93

@gleh Installed those RC packages in my test environment.
Tested basic operation. Looks good so far.
(Booted, all 3 SR connected, VMs able to start, both VHD and QCOW2)
Best regards

flakpyro

Installed on my usual hosts without issues. I am only making limited use of QCOW2 however.