A deployment kept failing with a quota error on a subscription that, by every dashboard I looked at, had plenty of room. The noise: three different numbers for "available cores" depending on where you looked.
The signal: regional quotas and VM-family quotas are separate ceilings, and the one that bites you is whichever is lower — per region, per family. The portal's headline number was the regional total; the deployment was blocked by the family-level limit in that one region.
The fix was a quota increase request scoped to the specific family and region, not the subscription. Ten minutes of work once I knew where to look. The lesson worth keeping: when a cloud limit makes no sense, you're almost always reading a different limit than the one being enforced.
Got a signal worth building around? Start a conversation →