# Short description
The Redis extended latency stats track per command latencies and enables:
- exporting the per-command percentile distribution via the `INFO LATENCYSTATS` command.
**( percentile distribution is not mergeable between cluster nodes ).**
- exporting the per-command cumulative latency distributions via the `LATENCY HISTOGRAM` command.
Using the cumulative distribution of latencies we can merge several stats from different cluster nodes
to calculate aggregate metrics .
By default, the extended latency monitoring is enabled since the overhead of keeping track of the
command latency is very small.
If you don't want to track extended latency metrics, you can easily disable it at runtime using the command:
- `CONFIG SET latency-tracking no`
By default, the exported latency percentiles are the p50, p99, and p999.
You can alter them at runtime using the command:
- `CONFIG SET latency-tracking-info-percentiles "0.0 50.0 100.0"`
## Some details:
- The total size per histogram should sit around 40 KiB. We only allocate those 40KiB when a command
was called for the first time.
- With regards to the WRITE overhead As seen below, there is no measurable overhead on the achievable
ops/sec or full latency spectrum on the client. Including also the measured redis-benchmark for unstable
vs this branch.
- We track from 1 nanosecond to 1 second ( everything above 1 second is considered +Inf )
## `INFO LATENCYSTATS` exposition format
- Format: `latency_percentiles_usec_<CMDNAME>:p0=XX,p50....`
## `LATENCY HISTOGRAM [command ...]` exposition format
Return a cumulative distribution of latencies in the format of a histogram for the specified command names.
The histogram is composed of a map of time buckets:
- Each representing a latency range, between 1 nanosecond and roughly 1 second.
- Each bucket covers twice the previous bucket's range.
- Empty buckets are not printed.
- Everything above 1 sec is considered +Inf.
- At max there will be log2(1000000000)=30 buckets
We reply a map for each command in the format:
`<command name> : { `calls`: <total command calls> , `histogram` : { <bucket 1> : latency , < bucket 2> : latency, ... } }`
Co-authored-by: Oran Agra <oran@redislabs.com>
Modules that use background threads with thread safe contexts are likely
to use RM_BlockClient() without a timeout function, because they do not
set up a timeout.
Before this commit, `CLIENT UNBLOCK` would result with a crash as the
`NULL` timeout callback is called. Beyond just crashing, this is also
logically wrong as it may throw the module into an unexpected client
state.
This commits makes `CLIENT UNBLOCK` on such clients behave the same as
any other client that is not in a blocked state and therefore cannot be
unblocked.
- removes time sensitive checks from block on background tests during leak checks.
- fix uninitialized variable on RedisModuleBlockedClient() when calling
RM_BlockedClientMeasureTimeEnd() without RM_BlockedClientMeasureTimeStart()
The test failed from time to time on Github actions.
We think it's possible that on the module's blocking timeout
time tracking test, the timeout is happening prior we issue the
RedisModule_BlockedClientMeasureTimeStart(bc) on the
background thread. If that is the case one possible solution
is to increase the timeout.
Increasing to 200ms to 500ms to see if nightly stops failing.
This commit enables tracking time of the background tasks and on replies,
opening the door for properly tracking commands that rely on blocking / background
work via the slowlog, latency history, and commandstats.
Some notes:
- The time spent blocked waiting for key changes, or blocked on synchronous
replication is not accounted for.
- **This commit does not affect latency tracking of commands that are non-blocking
or do not have background work.** ( meaning that it all stays the same with exception to
`BZPOPMIN`,`BZPOPMAX`,`BRPOP`,`BLPOP`, etc... and module's commands that rely
on background threads ).
- Specifically for latency history command we've added a new event class named
`command-unblocking` that will enable latency monitoring on commands that spawn
background threads to do the work.
- For blocking commands we're now considering the total time of a command as the
time spent on call() + the time spent on replying when unblocked.
- For Modules commands that rely on background threads we're now considering the
total time of a command as the time spent on call (main thread) + the time spent on
the background thread ( if marked within `RedisModule_MeasureTimeStart()` and
`RedisModule_MeasureTimeEnd()` ) + the time spent on replying (main thread)
To test for this feature we've added a `unit/moduleapi/blockonbackground` test that relies on
a module that blocks the client and sleeps on the background for a given time.
- check blocked command that uses RedisModule_MeasureTimeStart() is tracking background time
- check blocked command that uses RedisModule_MeasureTimeStart() is tracking background time even in timeout
- check blocked command with multiple calls RedisModule_MeasureTimeStart() is tracking the total background time
- check blocked command without calling RedisModule_MeasureTimeStart() is not reporting background time
