Эта страница ещё не переведена на язык «русский»: вот она на английском.
A plan’s limits are the same for every account on it. On Free, Paid raises most of them at once (glac billing plan paid), and the refusal of a call says so. Some of them we also raise for one account, when it needs more than its plan gives: a refusal that can be lifted this way says so in its message (“ask us to raise it for this account”), and its docs_url leads to its error. The account asks with one call of the API, below.
What we raise
| What | Refused with | What we do |
|---|---|---|
| What the account may have at once, and how fast it makes things: sandboxes and instances awake, their vCPUs and memory, sleeping sandboxes kept, an app’s instances in a region, sandboxes made a minute and a project’s made an hour, Cubes and versions made an hour | 403 plan_cap_reached, whose param names the limit |
Give the account a number of its own for any of them |
| How many Drives and volumes a project keeps, on Paid | 403 drive_limit_reached, volume_limit_reached |
Give the account a number of its own: on Paid it is a soft limit |
| How large a Drive or a volume may be, past every plan’s | The same | Let an enterprise account have Drives of up to 16 TiB |
| How much may be charged to the account’s card a month: its trust level | 403 trust_limit_exceeded |
Raise the level sooner than it rises by itself |
A number given to the account holds whatever plan it moves to, until it is taken back. GET /v1/account (glac account) shows each limit of what the account may have at once, what is used of it, and whose number it is (caps: source plan or operator), and what the key’s project keeps of Drives and volumes (keeps: how many, how large one may be, how large all of them may be together, each with what the project has, and the size of one made with no size); a refusal says “this account may” where the number is the account’s own.
How to ask
In the console, the account’s Limits page lists each of these limits with its number now, what is used of it and whose number it is; Ask for more on a limit’s row asks for it, with the number wanted and what for, and the requests are listed below it with their answers.
Or ask with the API, with a key that allows admin:
curl -X POST https://api.glac.ai/v1/limits/requests \
-H "Authorization: Bearer $GLAC_API_KEY" -H 'Content-Type: application/json' \
-d '{"limit": "vcpus", "wanted": 64, "why": "a build farm of eight sandboxes of 8 vCPUs"}'limit: the limit, by the name a refusal’sparamgives it, andGET /v1/accountundercaps:instances,vcpus,memory_bytes,sleeping,app_instances,creates_per_minute,creates_per_hour_per_projectorapp_creates_per_hour. Or what a project keeps:volumes(how many Drives and volumes) orvolume_bytes(how large one may be, up to 16 TiB).wanted: the number the account needs, more than it has now;memory_bytesandvolume_bytesin bytes.why: what it needs it for, 1 to 2 000 characters.
The answer is the request (limit_request, its id lrq-…), pending, with the number the account had when it asked (from). Asking changes nothing yet. One request of a limit is open at a time: another of the same limit is refused with 409 limit_request_pending, which names the one that waits. GET /v1/limits/requests lists the account’s requests, the newest first, and GET /v1/limits/requests/{id} reads one; every project’s keys see the same.
The answer. We grant it, and the request says granted with the number the account has from then on (granted; a Drive’s or a volume’s size in whole GiB), or we refuse it, and it says refused with why (answer). Either way the account is told by an event in each of its projects (limit_request.granted, limit_request.refused: GET /v1/events, and the webhooks that take them) and by mail to its address. A number granted is the account’s own, as in the table above.
For what the call does not name, such as a trust level sooner than it rises by itself, write to support@glac.ai with the account’s id (glac account, or id of GET /v1/account), the refused call’s code and request_id, and what you need.