This is in line with the v1.5 change in paradigm to not keep
chunk.Descs without chunks around after a series maintenance.
It's mainly motivated by avoiding excessive amounts of RAM usage
during crash recovery.
The code avoids to create memory time series with zero chunk.Descs as
that is prone to trigger weird effects. (Series maintenance would
archive series with zero chunk.Descs, but we cannot do that here
because the archive indices still have to be checked.)
The fpIter was kind of cumbersome to use and required a lock for each
iteration (which wasn't even needed for the iteration at startup after
loading the checkpoint).
The new implementation here has an obvious penalty in memory, but it's
only 8 byte per series, so 80MiB for a beefy server with 10M memory
time series (which would probably need ~100GiB RAM, so the memory
penalty is only 0.1% of the total memory need).
The big advantage is that now series maintenance happens in order,
which leads to the time between two maintenances of the same series
being less random. Ideally, after each maintenance, the next
maintenance would tackle the series with the largest number of
non-persisted chunks. That would be quite an effort to find out or
track, but with the approach here, the next maintenance will tackle
the series whose previous maintenance is longest ago, which is a good
approximation.
While this commit won't change the _average_ number of chunks
persisted per maintenance, it will reduce the mean time a given chunk
has to wait for its persistence and thus reduce the steady-state
number of chunks waiting for persistence.
Also, the map iteration in Go is non-deterministic but not truly
random. In practice, the iteration appears to be somewhat "bucketed".
You can often observe a bunch of series with similar duration since
their last maintenance, i.e. you see batches of series with similar
number of chunks persisted per maintenance. If that batch is
relatively young, a whole lot of series are maintained with very few
chunks to persist. (See screenshot in PR for a better explanation.)
In case the execution of all rules takes longer than the configured rule
evaluation interval, one or more iterations will be skipped. This needs
to be visible to the opterator.
This is needed for federating non-instance level metrics, so they don't
end up with the instance label of the prometheus target.
Also sort external labels, so label output order is consistent.
This is a fairly easy attempt to dynamically evict chunks based on the
heap size. A target heap size has to be set as a command line flage,
so that users can essentially say "utilize 4GiB of RAM, and please
don't OOM".
The -storage.local.max-chunks-to-persist and
-storage.local.memory-chunks flags are deprecated by this
change. Backwards compatibility is provided by ignoring
-storage.local.max-chunks-to-persist and use
-storage.local.memory-chunks to set the new
-storage.local.target-heap-size to a reasonable (and conservative)
value (both with a warning).
This also makes the metrics intstrumentation more consistent (in
naming and implementation) and cleans up a few quirks in the tests.
Answers to anticipated comments:
There is a chance that Go 1.9 will allow programs better control over
the Go memory management. I don't expect those changes to be in
contradiction with the approach here, but I do expect them to
complement them and allow them to be more precise and controlled. In
any case, once those Go changes are available, this code has to be
revisted.
One might be tempted to let the user specify an estimated value for
the RSS usage, and then internall set a target heap size of a certain
fraction of that. (In my experience, 2/3 is a fairly safe bet.)
However, investigations have shown that RSS size and its relation to
the heap size is really really complicated. It depends on so many
factors that I wouldn't even start listing them in a commit
description. It depends on many circumstances and not at least on the
risk trade-off of each individual user between RAM utilization and
probability of OOMing during a RAM usage peak. To not add even more to
the confusion, we need to stick to the well-defined number we also use
in the targeting here, the sum of the sizes of heap objects.
Currently, if a series stops to exist, its head chunk will be kept
open for an hour. That prevents it from being persisted. Which
prevents it from being evicted. Which prevents the series from being
archived.
Most of the time, once no sample has been added to a series within the
staleness limit, we can be pretty confident that this series will not
receive samples anymore. The whole chain as described above can be
started after 5m instead of 1h. In the relaxed case, this doesn't
change a lot as the head chunk timeout is only checked during series
maintenance, and usually, a series is only maintained every six
hours. However, there is the typical scenario where a large service is
deployed, the deoply turns out to be bad, and then it is deployed
again within minutes, and quite quickly the number of time series has
tripled. That's the point where the Prometheus server is stressed and
switches (rightfully) into rushed mode. In that mode, time series are
processed as quickly as possible, but all of that is in vein if all of
those recently ended time series cannot be persisted yet for another
hour. In that scenario, this change will help most, and it's exactly
the scenario where help is most desperately needed.
Rationale: The default value for GOGC is 100, i.e. a garbage collected
is initialized once as many heap space has been allocated as was in
use after the last GC was done. This ratio doesn't make a lot of sense
in Prometheus, as typically about 60% of the heap is allocated for
long-lived memory chunks (most of which are around for many hours if
not days). Thus, short-lived heap objects are accumulated for quite
some time until they finally match the large amount of memory used by
bulk memory chunks and a gigantic GC cyle is invoked. With GOGC=40, we
are essentially reinstating "normal" GC behavior by acknowledging that
about 60% of the heap are used for long-term bulk storage.
The median Prometheus production server at SoundCloud runs a GC cycle
every 90 seconds. With GOGC=40, a GC cycle is run every 35 seconds
(which is still not very often). However, the effective RAM usage is
now reduced by about 30%. If settings are updated to utilize more RAM,
the time between GC cycles goes up again (as the heap size is larger
with more long-lived memory chunks, but the frequency of creating
short-lived heap objects does not change). On a quite busy large
Prometheus server, the timing changed from one GC run every 20s to one
GC run every 12s.
In the former case (just changing GOGC, leave everything else as it
is), the CPU usage increases by about 10% (on a mid-size referenc
server from 8.1 to 8.9). If settings are adjusted, the CPU
consumptions increases more drastically (from 8 cores to 13 cores on a
large reference server), despite GCs happening more rarely, presumably
because a 50% larger set of memory chunks is managed now. Having more
memory chunks is good in many regards, and most servers are running
out of memory long before they run out of CPU cycles, so the tradeoff
is overwhelmingly positive in most cases.
Power users can still set the GOGC environment variable as usual, as
the implementation in this commit honors an explicitly set variable.
Julius Volz