IT Consultant | Veeam Certified Engineer | Veeam Legend | VMware Certified Professional | VMware vExpert with 26 years of experience in system and network administration writing about backup, virtualisation, VMware, Veeam, and whatever else catches my eye. This blog has a mirror on Medium.
For further info and contacts, visit my personal site: www.tinivelli.com .
Where it came from Last week I was working with some colleagues on a migration from one Veeam Backup for Microsoft 365 server to another — the kind of job where most of the effort goes into making sure nothing gets left behind. While we were putting together a script to help with the move, something else came to mind: certificates. Every organization in VB365 authenticates against Microsoft 365 with a certificate, and certificates expire. Nothing dramatic happens on the day: the backups simply start failing, which is arguably worse, because you find out when you go looking. And when you have a good number of tenants, opening each app registration in Entra ID to check an expiry date is exactly the kind of chore nobody remembers to do. ...
The common pitfall A few weeks ago I spent a few days on-site at a customer running a pretty serious infrastructure: lots of vSphere hosts with tons of VMs, external customers, OpenShift, block storage backing the VMs, and NAS storage holding and serving some very critical production data. The infrastructure is split across two active-active sites. I want to focus on the NAS boxes, or filers, as we used to call them a long time ago. The vendor provides clustering and mirroring features. The architecture, quite rightly, is built around an active-active cluster that’s always in sync, with one node at each active site. For an extra layer of data protection, there’s also a mirror to a third box sitting in a different location from the first two. ...
Let me start with something I often tell clients, and that’s worth repeating to myself too: a single copy is not a backup, and a copy that lives only on someone else’s cloud isn’t really your copy either. GitHub is wonderfully convenient — I use it every day — but it’s a service, not an archive. And like any service, it can be pulled out from under you in ways that have nothing to do with you: ...
As is I recently bought my second Arduino Uno Q. This time I installed it as a “normal computer”, I mean I connected it to a USB hub, where an HDMI cable connected to my monitor and a network cable connected to my server network converge, together with a USB Power cable. And a wired keyboard and mouse too, just for the first install, soon replaced by bluetooth mouse and keyboard. ...
Easy but good I have been administering vSphere infrastructures for a long time, but most of them were in little/middle companies with standard licenses that did not allow using Distributed Virtual Switches. The default load balancing in vmkernels and Port Group is: Route based on originating virtual port In short terms, every VM is mapped to a physical NIC and moved to another one in case of failures only. It’s simple, does not require configuration on physical switches, and works pretty well… but, easy to understand, the load balancing between NICs is not optimal. When we were working with a lot of NICs, maybe 4 or more, because we had 1 Gbit/s only, it was not such a great problem, for statistical reasons. Now we often have only 2 10Gig NICs (or faster) and that primitive load balancing can cause issues that are easy to solve with the Distributed Virtual Switch features. ...
Enterprise Manager operations Adding the newly deployed VSA 13.1 to the Enterprise Manager was totally straight forward: click Add… and follow the wizard ONE operations adding correctly the two different (windows and vsa) backup server by adding the Enterprise Manager directly to ONE required some trick: first of all click on Data source -> Add server… and select Enterprise Manager in the wizard ...
Broadcom is no longer as friendly as VMware was vSphere (VVF or VCF) licenses are getting more expensive every month, and many companies decided to migrate to another hypervisor: ProxMox, Oracle Virtualization, Nutanix, Microsoft and HPE are only some examples of the market movers. Most, but not all, of them are KVM based. Some good guys developed a “product” that has a lot of nice features to help massive moving of VM to another hypervisor. At the moment only a few of them are supported, mainly ProxMox and HPE. Currently it is in beta and totally not official, not supported. But it looks good 🙂 ...
Freedom from HW vendors The story Veeam has recently acquired Object First but the v13.1 brings another feature that provides us more freedom of choice for the backup storage. It’s possible to (easily) attach LUNs from any block storage device, iSCSI or FCP. I gave it a try using an iSCSI LUN from a virtual Truenas server. I did not use any auth method but mapped the lun to the veeam VSA initiator only (see below how to find it) ...
Another episode in the fight DBAs vs Backup Admins? ☺️ A feature that landed recently, and that I find super useful in enterprise contexts, is granular permissions on backup jobs. Easy example: you give the DBAs the ability to manage the jobs for “their own” servers. This lets you avoid those situations we’ve all seen way too many times, where DB servers end up protected badly, or in some needlessly convoluted way — like having Veeam (or something similar) do only crash-consistent backups and then hoping — fingers crossed — that data consistency is being handled properly over on the DB team’s side. ...
Version 13.1 seems to be shipping so many things that, joking, should have been called 13.5 or even more ☺️ Facts I have found the Observability page in the VSA web interface -> Configuration page (not in the Host Management Control) and activated it https://helpcenter.veeam.com/docs/vbr/userguide/configure_node_exporter.html?ver=13 the output is not human readable 😃 but it uses a common format: Prometheus. This gives us a lot of opportunities ...