News, tutorials and feedback on the cloud and infrastructure IT
Pointing Glance, Cinder and Nova at one Ceph cluster is the easy part. Five failures survive a successful deployment, and none of them raises an …
Read articleRAG, LoRA, QLoRA ou fine-tuning complet : ce qui décide réellement le coût et la VRAM nécessaire pour adapter un LLM à vos données.
Read articleL'échec le plus fréquent n'est pas technique. Ce qui structure vraiment la décision, le coût que tout le monde oublie, et les quatre moments où …
Read articleA fresh kubeadm cluster enforces nothing: containers run as root and nothing warns you. How to measure the gap and close it without breaking production.
Read articleUne facture émise sans écriture comptable, avec un message de succès. Ce qu'un statut brouillon ajouté après coup révèle d'une application de gestion.
Read articleUne facture de 14 250 EUR comptabilisée comme 14 250 MAD. Pourquoi le sens d'un taux doit s'écrire, et les huit garde-fous qui l'imposent.
Read articleConverting the disk took ten seconds. Adapting the guest took five minutes. What a cold migration really consists of, and the three things that break …
Read articleWhy your datastore does not shrink after a prune, how the keep options really combine, and three documented traps that cost a restore.
Read articleProxmox now has a cross-cluster management layer. What PDM 1.1 actually covers, the five things it does not, and how to attach your first cluster.
Read articleUsable capacity is not raw capacity divided by three. The rules, the arithmetic and the five sizing mistakes that cost the most on a production …
Read articleJoin our newsletter to receive our latest technical items directly in your mailbox.