Lambda sold-out GPU tiers and stray instances: check capacity and list your pods

Lambda sold-out GPU tiers and stray instances: check capacity and list your pods

TL;DR

If Lambda reports no capacity for your GPU type in your region, do not keep hammering launch, widen your preferred instance types or retry later, tiers sell out. Audit running instances regularly with a list command and stop the ones you forgot, there is no auto-stop to save you. And if GitHub auth fails from inside the instance, check that your local ssh-agent is running and the key is added before you blame the instance.

Use this when you run into the situation in the title.

When not to use this skill: unrelated tasks. It covers only the procedure above.

Compatibility

The steps above apply to the commands named in them. This skill does not pin a version, so if a flag looks different on your machine, check your installed version's docs first.

Details

Context: Web (inoculate-or-reflect infra/lambda README, troubleshooting): documents real capacity and hygiene problems. find-capacity can fail for every preferred GPU type when Lambda is sold out of that tier in your filesystem's region, the fix is retrying later or widening the preferred instance types. A pod can be running without you realizing it, burning money, so list and stop stray instances. SSH agent forwarding to GitHub needs a running ssh-agent with the key added, passphrase-protected keys may need an interactive prompt the first time.

Source: Web (inoculate-or-reflect infra/lambda README, troubleshooting): documents real capacity and hygiene problems. find-capacity can fail for every preferred GPU type when Lambda is sold out of that tier in your filesystem's region, the fix is retrying later or widening the preferred instance types. A pod can be running without you realizing it, burning money, so list and stop stray instances. SSH agent forwarding to GitHub needs a running ssh-agent with the key added, passphrase-protected keys may need an interactive prompt the first time.

Verified resolution: If Lambda reports no capacity for your GPU type in your region, do not keep hammering launch, widen your preferred instance types or retry later, tiers sell out. Audit running instances regularly with a list command and stop the ones you forgot, there is no auto-stop to save you. And if GitHub auth fails from inside the instance, check that your local ssh-agent is running and the key is added before you blame the instance.

Source URL: https://github.com/ayesha-imr/inoculate-or-reflect/blob/HEAD/infra/lambda/README.md