This commit moves Scraper handling into a separate scrapePool type.
TargetSets only manage TargetProvider lifecycles and sync the
retrieved updates to the scrapePool.
TargetProviders are now expected to send a full initial target set
within 5 seconds. The scrapePools preserve target state across reloads
and only drop targets after the initial set was synced.
We group providers by their scrape configuration. Each provider produces
target groups with an unique identifier.
On stopping a set of target providers we cancel the target providers,
stop scraping the targets and wait for the scrapers to finish.
On configuration reload all provider sets are stopped and new ones
are created. This will make targets disappear briefly on configuration
reload. Potentially scrapes are missed but due to the consistent
scrape intervals implemented recently, the impact is minor.
Double acquisition of the RLock usually doesn't blow up, but if the
write lock is called for between the two RLock's, we are deadlocked.
This deadlock does not exist in release-0.17, BTW.
With recent changes to a Target's internal data representation
updating by fullLabels() assigns the additional default
instance label. This breaks target identity comparison and causes
identical targets from service discovery to be constantly swapped.
This will fix issue #1035 and will also help to make issue #1264 less
bad.
The fundamental problem in the current code:
In the preload phase, we quite accurately determine which chunks will
be used for the query being executed. However, in the subsequent step
of creating series iterators, the created iterators are referencing
_all_ in-memory chunks in their series, even the un-pinned ones. In
iterator creation, we copy a pointer to each in-memory chunk of a
series into the iterator. While this creates a certain amount of
allocation churn, the worst thing about it is that copying the chunk
pointer out of the chunkDesc requires a mutex acquisition. (Remember
that the iterator will also reference un-pinned chunks, so we need to
acquire the mutex to protect against concurrent eviction.) The worst
case happens if a series doesn't even contain any relevant samples for
the query time range. We notice that during preloading but then we
will still create a series iterator for it. But even for series that
do contain relevant samples, the overhead is quite bad for instant
queries that retrieve a single sample from each series, but still go
through all the effort of series iterator creation. All of that is
particularly bad if a series has many in-memory chunks.
This commit addresses the problem from two sides:
First, it merges preloading and iterator creation into one step,
i.e. the preload call returns an iterator for exactly the preloaded
chunks.
Second, the required mutex acquisition in chunkDesc has been greatly
reduced. That was enabled by a side effect of the first step, which is
that the iterator is only referencing pinned chunks, so there is no
risk of concurrent eviction anymore, and chunks can be accessed
without mutex acquisition.
To simplify the code changes for the above, the long-planned change of
ValueAtTime to ValueAtOrBefore time was performed at the same
time. (It should have been done first, but it kind of accidentally
happened while I was in the middle of writing the series iterator
changes. Sorry for that.) So far, we actively filtered the up to two
values that were returned by ValueAtTime, i.e. we invested work to
retrieve up to two values, and then we invested more work to throw one
of them away.
The SeriesIterator.BoundaryValues method can be removed once #1401 is
fixed. But I really didn't want to load even more changes into this
PR.
Benchmarks:
The BenchmarkFuzz.* benchmarks run 83% faster (i.e. about six times
faster) and allocate 95% fewer bytes. The reason for that is that the
benchmark reads one sample after another from the time series and
creates a new series iterator for each sample read.
To find out how much these improvements matter in practice, I have
mirrored a beefy Prometheus server at SoundCloud that suffers from
both issues #1035 and #1264. To reach steady state that would be
comparable, the server needs to run for 15d. So far, it has run for
1d. The test server currently has only half as many memory time series
and 60% of the memory chunks the main server has. The 90th percentile
rule evaluation cycle time is ~11s on the main server and only ~3s on
the test server. However, these numbers might get much closer over
time.
In addition to performance improvements, this commit removes about 150
LOC.
So far we were using the InstanceIdentifier to compare equality of targets.
This is not always accurate, for example for the blackbox exporter where the
actual target is in the parameter.
For historic reasons we were enforcing a timeout directly
via the TCP dialer. This is no longer necessary for quite a while now.
Switching to context.Context will allow us to properly terminate
requests on shutdown as well.
