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.
Also, in that code path, set chunkDescsOffset to 0 rather than -1 in
case of "dropped more chunks from persistence than from memory" so
that no other weird things happen before the series is quarantined for
good.
Add metrics around checkpointing and persistence
* Add a metric to say if checkpointing is happening,
and another to track total checkpoint time and count.
This breaks the existing prometheus_local_storage_checkpoint_duration_seconds
by renaming it to prometheus_local_storage_checkpoint_last_duration_seconds
as the former name is more appropriate for a summary.
* Add metric for last checkpoint size.
* Add metric for series/chunks processed by checkpoints.
For long checkpoints it'd be useful to see how they're progressing.
* Add metric for dirty series
* Add metric for number of chunks persisted per series.
You can get the number of chunks from chunk_ops,
but not the matching number of series. This helps determine
the size of the writes being made.
* Add metric for chunks queued for persistence
Chunks created includes both chunks that'll need persistence
and chunks read in for queries. This only includes chunks created
for persistence.
* Code review comments on new persistence metrics.
Two cases:
- An unarchived metric must have at least one chunk desc loaded upon
unarchival. Otherwise, the file is gone or has size 0, which is an
inconsistency (because the series is still indexed in the archive
index). Hence, quarantining is triggered.
- If loading the chunk descs of a series with a known chunkDescsOffset
(i.e. != -1), the number of chunks loaded must be equal to
chunkDescsOffset. If not, there is a data corruption. An error is
returned, which leads to qurantining.
In any case, there is a guard added to not access the 1st element of
an empty chunkDescs slice. (That's what triggered the crashes in issue
2249.) A time series with unknown chunkDescsOffset and no chunks in
memory and no chunks on disk either could trigger that case. I would
assume such a "null series" doesn't exist, but it's not entirely
unthinkable and unreasonable to happen (perhaps in future uses of the
storage). (Create a series, and then something tries to preload chunks
before the first sample is added.)
This extracts Querier as an instantiateable and closeable object
rather than just defining extending methods of the storage interface.
This improves composability and allows abstracting query transactions,
which can be useful for transaction-level caches, consistent data views,
and encapsulating teardown.
These more specific methods have replaced `metricForLabelMatchers`
in cases where its `map[fingerprint]metric` result type was
not necessary or was used as an intermediate step
Avoids duplicated calls to `seriesForRange` from
`QueryRange` and `QueryInstant` methods.
This is a followup to https://github.com/prometheus/prometheus/pull/2011.
This publishes more of the methods and other names of the chunk code and
moves the chunk code to its own package. There's some unavoidable
ugliness: the chunk and chunkDesc metrics are used by both packages, so
I had to move them to the chunk package. That isn't great, but I don't
see how to do it better without a larger redesign of everything. Same
for the evict requests and some other types.
This is based on https://github.com/prometheus/prometheus/pull/1997.
This adds contexts to the relevant Storage methods and already passes
PromQL's new per-query context into the storage's query methods.
The immediate motivation supporting multi-tenancy in Frankenstein, but
this could also be used by Prometheus's normal local storage to support
cancellations and timeouts at some point.
See discussion in
https://groups.google.com/forum/#!topic/prometheus-developers/bkuGbVlvQ9g
The main idea is that the user of a storage shouldn't have to deal with
fingerprints anymore, and should not need to do an individual preload
call for each metric. The storage interface needs to be made more
high-level to not expose these details.
This also makes it easier to reuse the same storage interface for remote
storages later, as fewer roundtrips are required and the fingerprint
concept doesn't work well across the network.
NOTE: this deliberately gets rid of a small optimization in the old
query Analyzer, where we dedupe instants and ranges for the same series.
This should have a minor impact, as most queries do not have multiple
selectors loading the same series (and at the same offset).
tl;dr: This is not a fundamental solution to the indexing problem
(like tindex is) but it at least avoids utilizing the intersection
problem to the greatest possible amount.
In more detail:
Imagine the following query:
nicely:aggregating:rule{job="foo",env="prod"}
While it uses a nicely aggregating recording rule (which might have a
very low cardinality), Prometheus still intersects the low number of
fingerprints for `{__name__="nicely:aggregating:rule"}` with the many
thousands of fingerprints matching `{job="foo"}` and with the millions
of fingerprints matching `{env="prod"}`. This totally innocuous query
is dead slow if the Prometheus server has a lot of time series with
the `{env="prod"}` label. Ironically, if you make the query more
complicated, it becomes blazingly fast:
nicely:aggregating:rule{job=~"foo",env=~"prod"}
Why so? Because Prometheus only intersects with non-Equal matchers if
there are no Equal matchers. That's good in this case because it
retrieves the few fingerprints for
`{__name__="nicely:aggregating:rule"}` and then starts right ahead to
retrieve the metric for those FPs and checking individually if they
match the other matchers.
This change is generalizing the idea of when to stop intersecting FPs
and go into "retrieve metrics and check them individually against
remaining matchers" mode:
- First, sort all matchers by "expected cardinality". Matchers
matching the empty string are always worst (and never used for
intersections). Equal matchers are in general consider best, but by
using some crude heuristics, we declare some better than others
(instance labels or anything that looks like a recording rule).
- Then go through the matchers until we hit a threshold of remaining
FPs in the intersection. This threshold is higher if we are already
in the non-Equal matcher area as intersection is even more expensive
here.
- Once the threshold has been reached (or we have run out of matchers
that do not match the empty string), start with "retrieve metrics
and check them individually against remaining matchers".
A beefy server at SoundCloud was spending 67% of its CPU time in index
lookups (fingerprintsForLabelPairs), serving mostly a dashboard that
is exclusively built with recording rules. With this change, it spends
only 35% in fingerprintsForLabelPairs. The CPU usage dropped from 26
cores to 18 cores. The median latency for query_range dropped from 14s
to 50ms(!). As expected, higher percentile latency didn't improve that
much because the new approach is _occasionally_ running into the worst
case while the old one was _systematically_ doing so. The 99th
percentile latency is now about as high as the median before (14s)
while it was almost twice as high before (26s).
PromQL only requires a much narrower interface than local.Storage in
order to run queries. Narrower interfaces are easier to replace and
test, too.
We could also change the web interface to use local.Querier, except that
we'll probably use appending functions from there in the future.
With a lot of series accessed in a short timeframe (by a query, a
large scrape, checkpointing, ...), there is actually quite a
significant amount of lock contention if something similar is running
at the same time.
In those cases, the number of locks needs to be increased.
On the same front, as our fingerprints don't have a lot of entropy, I
introduced some additional shuffling. With the current state, anly
changes in the least singificant bits of a FP would matter.
But only on DEBUG level.
Also, count and report the two cases of out-of-order timestamps on the
one hand and same timestamp but different value on the other hand
separately.
Before, we checkpointed after every newly detected fingerprint
collision, which is not a problem as long as collisions are
rare. However, with a sufficient number of metrics or particular
nature of the data set, there might be a lot of collisions, all to be
detected upon the first set of scrapes, and then the checkpointing
after each detection will take a quite long time (it's O(n²),
essentially).
Since we are rebuilding the fingerprint mapping during crash recovery,
the previous, very conservative approach didn't even buy us
anything. We only ever read from the checkpoint file after a clean
shutdown, so the only time we need to write the checkpoint file is
during a clean shutdown.
It is now also used in label matching, so the name of the metric
changed from `prometheus_local_storage_invalid_preload_requests_total`
to `non_existent_series_matches_total'.
Only return an error where callers are doing something with it except
simply logging and ignoring.
All the errors touched in this commit flag the storage as dirty
anyway, and that fact is logged anyway. So most of what is being
removed here is just log spam.
As discussed earlier, the class of errors that flags the storage as
dirty signals fundamental corruption, no even bubbling up a one-time
warning to the user (e.g. about incomplete results) isn't helping much
because _anything_ happening in the storage has to be doubted from
that point on (and in fact retroactively into the past, too). Flagging
the storage dirty, and alerting on it (plus marking the state in the
web UI) is the only way I can see right now.
As a byproduct, I cleaned up the setDirty method a bit and improved
the logged errors.
