-
-
Notifications
You must be signed in to change notification settings - Fork 38
Description
Description
Sometime a resource (as in lock management) can be split into sub-part. For example, Cloud Foundry quota, nodes in a distributed architecture, ...
So, it will be nice to support easily such kind of lock.
Proposal
Solution 1 : Pool metadata
Pool directory may contain metadata which describe pool & lock capacities. When acquired, lock weight may be specified and recorded into new file. When released, lock weight is read from same file.
Note: When lock weight isn't specified, all capacity is assumed.
Solution 2 : Multi acquire/release (simplest)
Users have to create enough locks as pool capacity. acquire now accepts a number (or adds acquire_count). A new count file is generated providing lock count (and/or a names file with lock names). Then, name.<index> and metadata.<index> are generated as existing.
release doesn't change signature but support new directory structure.
Workaround
Currently I use as enough lock as necessary.
But need:
- to split into a reasonable amount of locks. For example, I divided my 10240 MB Cloud Foundry quota into 40 slots of 256 MB ; while static web component only requires 64 MB.
- to add enough
putto acquire/release (not also being sure how to arrange release in order to be sure they are all correctly release in all cases)