This can happen in the situation where the system scales up the number of shards massively (to deal with some backlog), then scales it down again as the number of samples sent during the time period is less than the number received.
We would overscan when hitting a value directly, interspersed with
samples in between timestamps. Apparently, that happens rarely enough
that it was only noticed recently.
* Compress remote storage requests and responses with unframed/raw snappy, for compatibility with other languages.
* Remove backwards compatibility code from remote_storage_adapter, update example_write_adapter
* Add /documentation/examples/remote_storage/example_write_adapter/example_writer_adapter to .gitignore
* Use request.Context() instead of a global map of contexts.
* Add some basic opentracing instrumentation on the query path.
* Remove tracehandler endpoint.
- checkpointSeriesMapAndHeads accepts a context now to allow
cancelling.
- If a shutdown is initiated, cancel the ongoing checkpoint. (We will
create a final checkpoint anyway.)
- Always wait for at least as long as the last checkpoint took before
starting the next checkpoint (to cap the time spending checkpointing
at 50%).
- If an error has occurred during checkpointing, don't bother to sync
the write.
- Make sure the temporary checkpoint file is deleted, even if an error
has occurred.
- Clean up the checkpoint loop a bit. (The concurrent Timer.Reset(0)
call might have cause a race.)
Fixes#2480. For certain definition of "fixes".
This is something that should never happen. Sadly, it does happen,
albeit extremely rarely. This could be some weird cornercase we
haven't covered yet. Or it happens as a consequesnce of data
corruption or a crash recovery gone bad.
This is not a "real" fix as we don't know the root cause of the
incident reported in #2480. However, this makes sure the server does
not crash, but deals gracefully with the problem: The series in
question is quarantined, which even makes it available for forensics.
An unopenable archived_fingerprint_to_timerange is simply deleted and
will be rebuilt during crash recovery (wich can then take quite some time).
An unopenable archived_fingerprint_to_metric is not deleted but
instructions to the user are logged. A deletion has to be done by the
user explicitly as it means losing all archived series (and a repair
with a 3rd party tool might still be possible).
Sadly, we have a number of places where we use varint encoding for
numbers that cannot be negative. We could have saved a bit by using
uvarint encoding. On the bright side, we now have a 50% chance to
detect data corruption. :-/
Fixes#1800 and #2492.
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.)
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.
