vaultaire
vaultaire copied to clipboard
Data vault for metrics
We're putting fixing this on hold due to other priorities, but it should be documented. Currently, a batch read will enumerate the buckets it hits `O(n)` times in the number...
The `cleanup` actions in TestHelpers are utterly destructive, meaning two test programs can't run on the same pool at the same time as one would blow away the in-progress work...
Something about ff7bf80d0f009833e5ef682238cf4bcf2b9ab85d has caused the writer daemon to deadlock (?). Maybe it's only getting one capability, thereby messing up the multi-threaded runtime... but that's not supposed to happen. AfC
This is a more sensible default because you need to actually try to end up with zillions of independent telemetry time series aggregated under the same address (as the only...
It's unintuitive to require runtime configuration with both CLI options and envvars (particularly as the latter aren't mentioned in the output of `vault --help`). Work out how to specify the...
This is currently difficult to do without writing a custom application for it; all existing utilities treat sourcedicts as append-only (semantically, not just under the hood). I propose adding a...
Should take an origin and return something like `LastUpdatedResponse = LastUpdated TimeStamp | UnknownOrigin`. Reasoning for addition is so Chevalier/other things which need an up-to-date contents listing aren't constantly reindexing...
The contents/reader daemon needs a way to cancel ongoing operations to avoid doing wasteful work. This would probably be implemented as new ZMQ messages types.
- Variable-length fields cause all sorts of problems in practice (most immediately, I can't implement fixed-width telemetry messages). - This doesn't limit practical applications; if you need more than 2**64...
The profiler currently reports the mean latency for ceph/reader/writer/contents. It is probably more reliable to use the median instead.