Recent content by hvisage

Standard SysAdmin best practises from my perspective: - tiny replaceable OS/root partitions - Data storage to be separate from root/os storage -- then moving it between faster(nvme zfs) vs redundant (read ceph) vs cheaper (HDD zfs) depending on the needs/requirements etc. -- Also different...

I'm working on IPv6 only setup in my clusters, and I prefer to use SIIT-DC to not have NAT44 type gateway "funs", I ran into the funs w.r.t. greylisting and SFP checks. Q1: Is there a method to have a specific netblock (the NAT46 block with the lower 32bits encoding the source IPv4)...

ran into a "fun" problem: I'm busy with a new deploy (using LXC template in PVE), and had /var/spool/pmg/ on a CEPHfs the upgrade to 9.1 failed with pmg-api (9.1.0)'s package doing a chgrp -R /var/spool/pmg/, and then errors on /var/spool/pmg/.lost+found chgrp: cannot read directory...

Intel(R) Celeron(R) CPU N3150 @ 1.60GHz after kernel 7.0 upgrades, I got hit by this GPU purging bug thingy ;( I've been hit with it before, can't recall what I changed back then. I had like 50% RAM free in a 16GB system, yeah seemed swap (on zfs) was used even though I had like 50%free ram...

Ah ! that is great to know - I'm using a tad... custom solution(s) so I'll try it out on test clusters

Hmm... this is still "outstanding" but it seems progress have been made recently I was testing this, and it seemed inter-PVE 9.1 this worked in the SDN setup...... except.. I had to interconnect to a Arista with 1st gen Jericho that doesn't support IPv6 VxLAN VTEPs ;( ;( With 10.6 that...

hmmm... yeah, getting in a situation where specifying namespaces for some VMs that needs more aggressive pruning (sflow/netflow type data) would've been a perfect solution, as there aren't tags nor other finer grained pruning on PBS available per se, so now extra and extra storage (even if...

yeah, but in a home situation where I already have a PBS that also sync off-site, I now need to setup something ssh or S3, while the PBS would've been a nice touch - Apple in anycase move away from bootable clones/bootable external storage (when the internal storage fails) so data file backups...

LXCs are backed up using a tar type backup, which is the "normal" PBS backup client with added extras and optimizations for LXCs as far as I can tell. thus doing a tar type backup of the home directories would be a great solution... yeah, the rest of the macOS (given the lock down setup since...

It can't be DNS, can it??? Oh yes, it WAS DNS! `pvesm status -storage local` took >120 seconds.... and it turned out a single node was pointing to a problematic resolver, and once the resolver was fixed, stats popped up !!! just leaving it here for whomever had this problem ;)

I've been trying to use FAI https://fai-project.org/FAIme/ to create cloud-init images, but ran into troubles (can't recall was 1.5years ago) but if somebody could make that "simple" for server type installation, it could handle about all the concerns above :(

seems the core had been implemented in 8.1 :)

yeah, this is a Kernel 6.7+ (Though some mention already somewhere in 6.6) problem that is only fixed in net-snmp 5.10pre1... guess why it'll be still some time before that log flooding will go away with some nice SNMP values reported after that

But how does that work for the virtio-net-pci ? ie. not the real/physicals

I'm wondering about this, how a config that contains the mappings, is not "consistent" for a live migration? ie. you'll be using the same configuration on both sides, and use that to do the same mapping, which from the sysadmin's perspective to be consistent per his/her configuration file. I'm...