FAQ – Veeam Backup & Replication VCSP

FAQ – Veeam Backup & Replication VCSP

Work forever in progress 🙂

Les backups d’Exchange sont assez difficiles vous n’ĂȘtes pas les seuls, en cause gĂ©nĂ©ralement la « stabilitĂ© » du failover cluster Exchange qui supportent mal les snapshot VMware.

Voici les tweaks possible : https://forums.veeam.com/veeam-backup-replication-f2/exchange-2010-dag-backup-t8329-30.html#p93106

Ensuite il y en a qui rajoute un nƓud passif et qui ne backup que le nƓud passif ou alors ils mettent des agents 😃 (plus de stunt de Snap VMware ahah) https://www.veeam.com/kb2463

Il y a deux façons d’avoir des rapports personnalisĂ©s :

ReFS 2016 & 2019 sont trĂšs stable et best practice. C’est dĂ©ployĂ© partout.

Quelques recommandations si besoin : https://bp.veeam.com/vbr/VBP/3_Build_structures/B_Veeam_Components/B_backup_repositories/block.html

Ce n’est pas nous le problùme pour XFS c’est le kernel Centos qui n’est pas à jour sur Reflink 😊

Since we added fast cloning support for XFS repositories in v10, this question keeps coming up: why do we only support Ubuntu for repository servers, and how long will support for other distributions remain experimental? The vast majority of those asking are of course RHEL/CentOS users, which together with Ubuntu are the distributions of choice for over three quarters of Veeam users. To put it short: our biggest concern with RHEL/CentOS 8 is that they use the kernel version is 4.18. Even if block cloning I/O controls aka reflink were introduced back in the kernel version 4.9, early adopters reported a number of performance challenges, and XFS went through some significant changes to address those. You can read some details in this article, but the main point is that this work was not fully completed until 5.4 kernel. At the same time, Veeam makes some of the heaviest use of reflink functionality among all applications on the market – which really pushes file system metadata handling logic to its limits. For example, virtually all block cloning issues with ReFS were due to sub-optimal handling of ReFS metadata, which is something Microsoft kept addressing with every update. So at this time, based on all these facts and our history with ReFS, we consider Ubuntu 20.04 the safest choice for Linux-based backup repositories with XFS fast cloning enabled, because it comes with 5.4 kernel. Previous kernel versions had some challenges that needed to be addressed – and we cannot know whether those will impact our customers or not. Basically, we just need more field validation: more Veeam users who are willing to put those older kernels through its paces to prove it. To be perfectly clear, we ourselves did not see any issues with RHEL/CentOS 8 based XFS repositories internally, but then again – it took us a very long time to reproduce those infamous ReFS issues even after many customers have already been struggling with them. So this time, we decided it is better to be safe than sorry again! Having said that, right now things looks pretty solid with XFS – I just did a full text search for XFS across all recent support cases, and there doesn't appear to be many mentions outside of Veeam Agent for Linux. Although to be fair, I also don't expect XFS adoption for backup repositories to be anywhere near ReFS yet.

Pas de support natif de SQL Server on Linux, il faut passer par des script pre et post backup.

le Quiescing et l’app aware sont des technologies permettant d’avoir une sauvegarde consistante au niveau base de donnĂ©es des VMs.

Du coup, quand il n’y a pas de base de donnĂ©es ou d’AD, Exchange, Sharepoint, SQL, Oracle, aucune des deux technologies sont nĂ©cessaires.

Dans le cadre d’une nĂ©cessitĂ© de backup consistant sur Windows, c’est l’application aware processing qui est nĂ©cessaire, avec un compte local admin + SA / sysadmin + ouverture de flux vers la VMs pour pouvoir prĂ©venir les VSS. Si c’est un linux, ce sera gĂ©nĂ©ralement des scripts pre et post backup pour envoyer des commandes Ă  la base de donnĂ©es et la rendre consistante.

Et le dernier moyen, c’est la Quiescing des VMware tools qui freeze les IOPS des machines linux le temps du snapshot. C’est rarement utilisĂ© car cela implique un stunt de la VMs.

Si tu as quand mĂȘme besoin de quiescing, je te conseil de faire un job spĂ©cifique pour ces quelques VMs linux pour ne pas stunt toutes tes VMs de ton job. A l’inverse l’APP Aware Processing est granulaire par VMs (tu peux faire un disable all et enable juste celle qui t’intĂ©resse)