part of https://github.com/neondatabase/neon/issues/7124
Changes
-------
This PR replaces the `EphemeralFile::write_blob`-specifc `struct Writer`
with re-use of `owned_buffers_io::write::BufferedWriter`.
Further, it restructures the code to cleanly separate
* the high-level aspect of EphemeralFile's write_blob / read_blk API
* the page-caching aspect
* the aspect of IO
* performing buffered write IO to an underlying VirtualFile
* serving reads from either the VirtualFile or the buffer if it hasn't
been flushed yet
* the annoying "feature" that reads past the end of the written range
are allowed and expected to return zeroed memory, as long as one remains
within one PAGE_SZ
## Problem
We are currently supporting two read paths. No bueno.
## Summary of changes
High level: use vectored read path to serve get page requests - gated by
`get_impl` config
Low level:
1. Add ps config, `get_impl` to specify which read path to use when
serving get page requests
2. Fix base cached image handling for the vectored read path. This was
subtly broken: previously we
would not mark keys that went past their cached lsn as complete. This is
a self standing change which
could be its own PR, but I've included it here because writing separate
tests for it is tricky.
3. Fork get page to use either the legacy or vectored implementation
4. Validate the use of vectored read path when serving get page requests
against the legacy implementation.
Controlled by `validate_vectored_get` ps config.
5. Use the vectored read path to serve get page requests in tests (with
validation).
## Note
Since the vectored read path does not go through the page cache to read
buffers, this change also amounts to a removal of the buffer page cache. Materialized page cache
is still used.
As part of https://github.com/neondatabase/neon/pull/7375 and to improve
the current vectored get implementation, we separate the missing key
error out. This also saves us several Box allocations in the get page
implementation.
## Summary of changes
* Create a caching field of layer traversal id for each of the layer.
* Remove box allocations for layer traversal id retrieval and implement
MissingKey error message as before. This should be a little bit faster.
* Do not format error message until `Display`.
* For in-mem layer, the descriptor is different before/after frozen. I'm
using once lock for that.
---------
Signed-off-by: Alex Chi Z <chi@neon.tech>
Before, we asserted that a layer would only be loaded by the timeline
that initially created it. Now, with the ancestor detach, we will want
to utilize remote copy as much as possible, so we will need to open
other timeline layers as our own.
Cc: #6994
#7030 introduced an annoying papercut, deeming a failure to acquire a
strong reference to `LayerInner` from `DownloadedLayer::drop` as a
canceled eviction. Most of the time, it wasn't that, but just timeline
deletion or tenant detach with the layer not wanting to be deleted or
evicted.
When a Layer is dropped as part of a normal shutdown, the `Layer` is
dropped first, and the `DownloadedLayer` the second. Because of this, we
cannot detect eviction being canceled from the `DownloadedLayer::drop`.
We can detect it from `LayerInner::drop`, which this PR adds.
Test case is added which before had 1 started eviction, 2 canceled. Now
it accurately finds 1 started, 1 canceled.
For "timeline ancestor merge" or "timeline detach," we need to "cut"
delta layers at particular LSN. The name "truncate" is not used as it
would imply that a layer file changes, instead of what happens: we copy
keys with Lsn less than a "cut point".
Cc: #6994
Add the "copy delta layer prefix" operation to DeltaLayerInner, re-using
some of the vectored read internals. The code is `cfg(test)` until it
will be used later with a more complete integration test.
## Problem
This trace is emitted whenever a vectored read touches the end of a
delta layer file. It's a perfectly normal case, but I expected it to be
more rare when implementing the code.
## Summary of changes
Demote log to debug.
Problem
Currently, we base our time based layer rolling decision on the last
time we froze a layer. This means that if we roll a layer and then go
idle for longer than the checkpoint timeout the next layer will be
rolled after the first write. This is of course not desirable.
Summary of changes
Record the timepoint of the first write to an open layer and use that
for time based layer rolling decisions. Note that I had to keep
`Timeline::last_freeze_ts` for the sharded tenant disk consistent lsn
skip hack.
Fixes#7241
## Problem
The vectored read path holds the layer map lock while visiting a
timeline.
## Summary of changes
* Rework the fringe order to hold `Layer` on `Arc<InMemoryLayer>`
handles instead of descriptions that are resolved by the layer map at
the time of read. Note that previously `get_values_reconstruct_data` was
implemented for the layer description which already knew the lsn range
for the read. Now it is implemented on the new `ReadableLayer` handle
and needs to get the lsn range as an argument.
* Drop the layer map lock after updating the fringe.
Related https://github.com/neondatabase/neon/issues/6833
## Problem
During this week's deployment we observed panics due to the blobs
for certain keys not fitting in the vectored read buffers. The likely
cause of this is a bloated AUX_FILE_KEY caused by logical replication.
## Summary of changes
This pr fixes the issue by allocating a buffer big enough to fit
the widest read. It also has the benefit of saving space if all keys
in the read have blobs smaller than the max vectored read size.
If the soft limit for the max size of a vectored read is violated,
we print a warning which includes the offending key and lsn.
A randomised (but deterministic) end to end test is also added for
vectored reads on the delta layer.
## Problem
Follows: https://github.com/neondatabase/neon/pull/7182
- Sufficient concurrent writes could OOM a pageserver from the size of
indices on all the InMemoryLayer instances.
- Enforcement of checkpoint_period only happened if there were some
writes.
Closes: https://github.com/neondatabase/neon/issues/6916
## Summary of changes
- Add `ephemeral_bytes_per_memory_kb` config property. This controls the
ratio of ephemeral layer capacity to memory capacity. The weird unit is
to enable making the ratio less than 1:1 (set this property to 1024 to
use 1MB of ephemeral layers for every 1MB of RAM, set it smaller to get
a fraction).
- Implement background layer rolling checks in
Timeline::compaction_iteration -- this ensures we apply layer rolling
policy in the absence of writes.
- During background checks, if the total ephemeral layer size has
exceeded the limit, then roll layers whose size is greater than the mean
size of all ephemeral layers.
- Remove the tick() path from walreceiver: it isn't needed any more now
that we do equivalent checks from compaction_iteration.
- Add tests for the above.
---------
Co-authored-by: Arpad Müller <arpad-m@users.noreply.github.com>
## Problem
Large quantities of ephemeral layer data can lead to excessive memory
consumption (https://github.com/neondatabase/neon/issues/6939). We
currently don't have a way to know how much ephemeral layer data is
present on a pageserver.
Before we can add new behaviors to proactively roll layers in response
to too much ephemeral data, we must calculate that total.
Related: https://github.com/neondatabase/neon/issues/6916
## Summary of changes
- Create GlobalResources and GlobalResourceUnits types, where timelines
carry a GlobalResourceUnits in their TimelineWriterState.
- Periodically update the size in GlobalResourceUnits:
- During tick()
- During layer roll
- During put() if the latest value has drifted more than 10MB since our
last update
- Expose the value of the global ephemeral layer bytes counter as a
prometheus metric.
- Extend the lifetime of TimelineWriterState:
- Instead of dropping it in TimelineWriter::drop, let it remain.
- Drop TimelineWriterState in roll_layer: this drops our guard on the
global byte count to reflect the fact that we're freezing the layer.
- Ensure the validity of the later in the writer state by clearing the
state in the same place we freeze layers, and asserting on the
write-ability of the layer in `writer()`
- Add a 'context' parameter to `get_open_layer_action` so that it can
skip the prev_lsn==lsn check when called in tick() -- this is needed
because now tick is called with a populated state, where
prev_lsn==Some(lsn) is true for an idle timeline.
- Extend layer rolling test to use this metric
Before this PR, each core had 3 executor threads from 3 different
runtimes. With this PR, we just have one runtime, with one thread per
core. Switching to a single tokio runtime should reduce that effective
over-commit of CPU and in theory help with tail latencies -- iff all
tokio tasks are well-behaved and yield to the runtime regularly.
Are All Tasks Well-Behaved? Are We Ready?
-----------------------------------------
Sadly there doesn't seem to be good out-of-the box tokio tooling to
answer this question.
We *believe* all tasks are well behaved in today's code base, as of the
switch to `virtual_file_io_engine = "tokio-epoll-uring"` in production
(https://github.com/neondatabase/aws/pull/1121).
The only remaining executor-thread-blocking code is walredo and some
filesystem namespace operations.
Filesystem namespace operations work is being tracked in #6663 and not
considered likely to actually block at this time.
Regarding walredo, it currently does a blocking `poll` for read/write to
the pipe file descriptors we use for IPC with the walredo process.
There is an ongoing experiment to make walredo async (#6628), but it
needs more time because there are surprisingly tricky trade-offs that
are articulated in that PR's description (which itself is still WIP).
What's relevant for *this* PR is that
1. walredo is always CPU-bound
2. production tail latencies for walredo request-response
(`pageserver_wal_redo_seconds_bucket`) are
- p90: with few exceptions, low hundreds of micro-seconds
- p95: except on very packed pageservers, below 1ms
- p99: all below 50ms, vast majority below 1ms
- p99.9: almost all around 50ms, rarely at >= 70ms
- [Dashboard
Link](https://neonprod.grafana.net/d/edgggcrmki3uof/2024-03-walredo-latency?orgId=1&var-ds=ZNX49CDVz&var-pXX_by_instance=0.9&var-pXX_by_instance=0.99&var-pXX_by_instance=0.95&var-adhoc=instance%7C%21%3D%7Cpageserver-30.us-west-2.aws.neon.tech&var-per_instance_pXX_max_seconds=0.0005&from=1711049688777&to=1711136088777)
The ones below 1ms are below our current threshold for when we start
thinking about yielding to the executor.
The tens of milliseconds stalls aren't great, but, not least because of
the implicit overcommit of CPU by the three runtimes, we can't be sure
whether these tens of milliseconds are inherently necessary to do the
walredo work or whether we could be faster if there was less contention
for CPU.
On the first item (walredo being always CPU-bound work): it means that
walredo processes will always compete with the executor threads.
We could yield, using async walredo, but then we hit the trade-offs
explained in that PR.
tl;dr: the risk of stalling executor threads through blocking walredo
seems low, and switching to one runtime cleans up one potential source
for higher-than-necessary stall times (explained in the previous
paragraphs).
Code Changes
------------
- Remove the 3 different runtime definitions.
- Add a new definition called `THE_RUNTIME`.
- Use it in all places that previously used one of the 3 removed
runtimes.
- Remove the argument from `task_mgr`.
- Fix failpoint usage where `pausable_failpoint!` should have been used.
We encountered some actual failures because of this, e.g., hung
`get_metric()` calls during test teardown that would client-timeout
after 300s.
As indicated by the comment above `THE_RUNTIME`, we could take this
clean-up further.
But before we create so much churn, let's first validate that there's no
perf regression.
Performance
-----------
We will test this in staging using the various nightly benchmark runs.
However, the worst-case impact of this change is likely compaction
(=>image layer creation) competing with compute requests.
Image layer creation work can't be easily generated & repeated quickly
by pagebench.
So, we'll simply watch getpage & basebackup tail latencies in staging.
Additionally, I have done manual benchmarking using pagebench.
Report:
https://neondatabase.notion.site/2024-03-23-oneruntime-change-benchmarking-22a399c411e24399a73311115fb703ec?pvs=4
Tail latencies and throughput are marginally better (no regression =
good).
Except in a workload with 128 clients against one tenant.
There, the p99.9 and p99.99 getpage latency is about 2x worse (at
slightly lower throughput).
A dip in throughput every 20s (compaction_period_ is clearly visible,
and probably responsible for that worse tail latency.
This has potential to improve with async walredo, and is an edge case
workload anyway.
Future Work
-----------
1. Once this change has shown satisfying results in production, change
the codebase to use the ambient runtime instead of explicitly
referencing `THE_RUNTIME`.
2. Have a mode where we run with a single-threaded runtime, so we
uncover executor stalls more quickly.
3. Switch or write our own failpoints library that is async-native:
https://github.com/neondatabase/neon/issues/7216
## Problem
The current implementation of struct Layer supports canceled read
requests, but those will leave the internal state such that a following
`Layer::keep_resident` call will need to repair the state. In
pathological cases seen during generation numbers resetting in staging
or with too many in-progress on-demand downloads, this repair activity
will need to wait for the download to complete, which stalls disk
usage-based eviction. Similar stalls have been observed in staging near
disk-full situations, where downloads failed because the disk was full.
Fixes#6028 or the "layer is present on filesystem but not evictable"
problems by:
1. not canceling pending evictions by a canceled
`LayerInner::get_or_maybe_download`
2. completing post-download initialization of the `LayerInner::inner`
from the download task
Not canceling evictions above case (1) and always initializing (2) lead
to plain `LayerInner::inner` always having the up-to-date information,
which leads to the old `Layer::keep_resident` never having to wait for
downloads to complete. Finally, the `Layer::keep_resident` is replaced
with `Layer::is_likely_resident`. These fix#7145.
## Summary of changes
- add a new test showing that a canceled get_or_maybe_download should
not cancel the eviction
- switch to using a `watch` internally rather than a `broadcast` to
avoid hanging eviction while a download is ongoing
- doc changes for new semantics and cleanup
- fix `Layer::keep_resident` to use just `self.0.inner.get()` as truth
as `Layer::is_likely_resident`
- remove `LayerInner::wanted_evicted` boolean as no longer needed
Builds upon: #7185. Cc: #5331.
Before this PR, cancellation for `LayerInner::get_or_maybe_download`
could occur so that we have downloaded the layer file in the filesystem,
but because of the cancellation chance, we have not set the internal
`LayerInner::inner` or initialized the state. With the detached init
support introduced in #7135 and in place in #7152, we can now initialize
the internal state after successfully downloading in the spawned task.
The next PR will fix the remaining problems that this PR leaves:
- `Layer::keep_resident` is still used because
- `Layer::get_or_maybe_download` always cancels an eviction, even when
canceled
Split off from #7030. Stacked on top of #7152. Cc: #5331.
The second part of work towards fixing `Layer::keep_resident` so that it
does not need to repair the internal state. #7135 added a nicer API for
initialization. This PR uses it to remove a few indentation levels and
the loop construction. The next PR #7175 will use the refactorings done
in this PR, and always initialize the internal state after a download.
Cc: #5331
Since #6115 with more often used get_value_reconstruct_data and friends,
we should not have needless INFO level span creation near hot paths. In
our prod configuration, INFO spans are always created, but in practice,
very rarely anything at INFO level is logged underneath.
`ResidentLayer::load_keys` is only used during compaction so it is not
that hot, but this aligns the access paths and their span usage.
PR changes the span level to debug to align with others, and adds the
layer name to the error which was missing.
Split off from #7030.
Split off from #7030:
- each early exit is counted as canceled init, even though it most
likely was just `LayerInner::keep_resident` doing the no-download repair
check
- `downloaded_after` could had been accounted for multiple times, and
also when repairing to match on-disk state
Cc: #5331
This is a follow-up to #7051 where `LayerInner::drop` and
`LayerInner::evict_blocking` were not noticed to require a gate before
the file deletion. The lack of entering a gate opens up a similar
possibility of deleting a layer file which a newer Timeline instance has
already checked out to be resident in a similar case as #7051.
## Problem
Before this PR, it was possible that on-demand downloads were started
after `Timeline::shutdown()`.
For example, we have observed a walreceiver-connection-handler-initiated
on-demand download that was started after `Timeline::shutdown()`s final
`task_mgr::shutdown_tasks()` call.
The underlying issue is that `task_mgr::shutdown_tasks()` isn't sticky,
i.e., new tasks can be spawned during or after
`task_mgr::shutdown_tasks()`.
Cc: https://github.com/neondatabase/neon/issues/4175 in lieu of a more
specific issue for task_mgr. We already decided we want to get rid of it
anyways.
Original investigation:
https://neondb.slack.com/archives/C033RQ5SPDH/p1709824952465949
## Changes
- enter gate while downloading
- use timeline cancellation token for cancelling download
thereby, fixes#7054
Entering the gate might also remove recent "kept the gate from closing"
in staging.
## Problem
We reverted https://github.com/neondatabase/neon/pull/6661 a few days
ago. The change led to OOMs in
benchmarks followed by large WAL reingests.
The issue was that we removed [this
code](d04af08567/pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs (L409-L417)).
That call may trigger a roll of the open layer due to
the keepalive messages received from the safekeeper. Removing it meant
that enforcing
of checkpoint timeout became even more lax and led to using up large
amounts of memory
for the in memory layer indices.
## Summary of changes
Piggyback on keep alive messages to enforce checkpoint timeout. This is
a hack, but it's exactly what
the current code is doing.
## Alternatives
Christhian, Joonas and myself sketched out a timer based approach
[here](https://github.com/neondatabase/neon/pull/6940). While discussing
it further, it became obvious that's also a bit of a hack and not the
desired end state. I chose not
to take that further since it's not what we ultimately want and it'll be
harder to rip out.
Right now it's unclear what the ideal system behaviour is:
* early flushing on memory pressure, or ...
* detaching tenants on memory pressure
## Problem
Collection of small changes, batched together to reduce CI overhead.
## Summary of changes
- Layer download messages include size -- this is useful when watching a
pageserver hydrate its on disk cache in the log.
- Controller migrate API could put an invalid NodeId into TenantState
- Scheduling errors during tenant create could result in creating some
shards and not others.
- Consistency check could give hard-to-understand failures in tests if a
reconcile was in process: explicitly fail the check if reconciles are in
progress instead.
## Problem
Last weeks enablement of vectored get generated a number of panics.
From them, I diagnosed two issues in the delta layer index traversal
logic
1. The `key >= range.start && lsn >= lsn_range.start`
was too aggressive. Lsns are not monotonically increasing in the delta
layer index (keys are though), so we cannot assert on them.
2. Lsns greater or equal to `lsn_range.end` were not skipped. This
caused the query to consider records newer than the request Lsn.
## Summary of changes
* Fix the issues mentioned above inline
* Refactor the layer traversal logic to make it unit testable
* Add unit test which reproduces the failure modes listed above.
part of #6663
See that epic for more context & related commits.
Problem
-------
Before this PR, the layer-file-creating code paths were using
VirtualFile, but under the hood these were still blocking system calls.
Generally this meant we'd stall the executor thread, unless the caller
"knew" and used the following pattern instead:
```
spawn_blocking(|| {
Handle::block_on(async {
VirtualFile::....().await;
})
}).await
```
Solution
--------
This PR adopts `tokio-epoll-uring` on the layer-file-creating code paths
in pageserver.
Note that on-demand downloads still use `tokio::fs`, these will be
converted in a future PR.
Design: Avoiding Regressions With `std-fs`
------------------------------------------
If we make the VirtualFile write path truly async using
`tokio-epoll-uring`, should we then remove the `spawn_blocking` +
`Handle::block_on` usage upstack in the same commit?
No, because if we’re still using the `std-fs` io engine, we’d then block
the executor in those places where previously we were protecting us from
that through the `spawn_blocking` .
So, if we want to see benefits from `tokio-epoll-uring` on the write
path while also preserving the ability to switch between
`tokio-epoll-uring` and `std-fs` , where `std-fs` will behave identical
to what we have now, we need to ***conditionally* use `spawn_blocking +
Handle::block_on`** .
I.e., in the places where we use that know, we’ll need to make that
conditional based on the currently configured io engine.
It boils down to investigating all the places where we do
`spawn_blocking(... block_on(... VirtualFile::...))`.
Detailed [write-up of that investigation in
Notion](https://neondatabase.notion.site/Surveying-VirtualFile-write-path-usage-wrt-tokio-epoll-uring-integration-spawn_blocking-Handle-bl-5dc2270dbb764db7b2e60803f375e015?pvs=4
), made publicly accessible.
tl;dr: Preceding PRs addressed the relevant call sites:
- `metadata` file: turns out we could simply remove it (#6777, #6769,
#6775)
- `create_delta_layer()`: made sensitive to `virtual_file_io_engine` in
#6986
NB: once we are switched over to `tokio-epoll-uring` everywhere in
production, we can deprecate `std-fs`; to keep macOS support, we can use
`tokio::fs` instead. That will remove this whole headache.
Code Changes In This PR
-----------------------
- VirtualFile API changes
- `VirtualFile::write_at`
- implement an `ioengine` operation and switch `VirtualFile::write_at`
to it
- `VirtualFile::metadata()`
- curiously, we only use it from the layer writers' `finish()` methods
- introduce a wrapper `Metadata` enum because `std::fs::Metadata` cannot
be constructed by code outside rust std
- `VirtualFile::sync_all()` and for completeness sake, add
`VirtualFile::sync_data()`
Testing & Rollout
-----------------
Before merging this PR, we ran the CI with both io engines.
Additionally, the changes will soak in staging.
We could have a feature gate / add a new io engine
`tokio-epoll-uring-write-path` to do a gradual rollout. However, that's
not part of this PR.
Future Work
-----------
There's still some use of `std::fs` and/or `tokio::fs` for directory
namespace operations, e.g. `std::fs::rename`.
We're not addressing those in this PR, as we'll need to add the support
in tokio-epoll-uring first. Note that rename itself is usually fast if
the directory is in the kernel dentry cache, and only the fsync after
rename is slow. These fsyncs are using tokio-epoll-uring, so, the impact
should be small.
Nightly has added a bunch of compiler and linter warnings. There is also
two dependencies that fail compilation on latest nightly due to using
the old `stdsimd` feature name. This PR fixes them.
Because of bugs evictions could hang and pause disk usage eviction task.
One such bug is known and fixed#6928. Guard each layer eviction with a
modest timeout deeming timeouted evictions as failures, to be
conservative.
In addition, add logging and metrics recording on each eviction
iteration:
- log collection completed with duration and amount of layers
- per tenant collection time is observed in a new histogram
- per tenant layer count is observed in a new histogram
- record metric for collected, selected and evicted layer counts
- log if eviction takes more than 10s
- log eviction completion with eviction duration
Additionally remove dead code for which no dead code warnings appeared
in earlier PR.
Follow-up to: #6060.
## Problem
The vectored read path proposed in
https://github.com/neondatabase/neon/pull/6576 seems
to be functionally correct, but in my testing (see below) it is about 10-20% slower than the naive
sequential vectored implementation.
## Summary of changes
There's three parts to this PR:
1. Supporting vectored blob reads. This is actually trickier than it
sounds because on disk blobs are prefixed with a variable length size header.
Since the blobs are not necessarily fixed size, we need to juggle the offsets
such that the callers can retrieve the blobs from the resulting buffer.
2. Merge disk read requests issued by the vectored read path up to a
maximum size. Again, the merging is complicated by the fact that blobs
are not fixed size. We keep track of the begin and end offset of each blob
and pass them into the vectored blob reader. In turn, the reader will return
a buffer and the offsets at which the blobs begin and end.
3. A benchmark for basebackup requests against tenant with large SLRU
block counts is added. This required a small change to pagebench and a new config
variable for the pageserver which toggles the vectored get validation.
We can probably optimise things further by adding a little bit of
concurrency for our IO. In principle, it's as simple as spawning a task which deals with issuing
IO and doing the serialisation and handling on the parent task which receives input via a
channel.
This reverts commits 587cb705b8 (PR #6661)
and fcbe9fb184 (PR #6842).
Conflicts:
pageserver/src/tenant.rs
pageserver/src/tenant/timeline.rs
The conflicts were with
* pageserver: adjust checkpoint distance for sharded tenants (#6852)
* pageserver: add vectored get implementation (#6576)
Also we had to keep the `allowed_errors` to make `test_forward_compatibility` happy,
see the PR thread on GitHub for details.
Not allowing evicting wanted deleted layers is something I've forgotten
to implement on #5645. This PR makes it possible to evict such layers,
which should reduce the amount of hanging evictions.
Fixes: #6928
Co-authored-by: Christian Schwarz <christian@neon.tech>
Rebased version of #5234, part of #6768
This consists of three parts:
1. A refactoring and new contract for implementing and testing
compaction.
The logic is now in a separate crate, with no dependency on the
'pageserver' crate. It defines an interface that the real pageserver
must implement, in order to call the compaction algorithm. The interface
models things like delta and image layers, but just the parts that the
compaction algorithm needs to make decisions. That makes it easier unit
test the algorithm and experiment with different implementations.
I did not convert the current code to the new abstraction, however. When
compaction algorithm is set to "Legacy", we just use the old code. It
might be worthwhile to convert the old code to the new abstraction, so
that we can compare the behavior of the new algorithm against the old
one, using the same simulated cases. If we do that, have to be careful
that the converted code really is equivalent to the old.
This inclues only trivial changes to the main pageserver code. All the
new code is behind a tenant config option. So this should be pretty safe
to merge, even if the new implementation is buggy, as long as we don't
enable it.
2. A new compaction algorithm, implemented using the new abstraction.
The new algorithm is tiered compaction. It is inspired by the PoC at PR
#4539, although I did not use that code directly, as I needed the new
implementation to fit the new abstraction. The algorithm here is less
advanced, I did not implement partial image layers, for example. I
wanted to keep it simple on purpose, so that as we add bells and
whistles, we can see the effects using the included simulator.
One difference to #4539 and your typical LSM tree implementations is how
we keep track of the LSM tree levels. This PR doesn't have a permanent
concept of a level, tier or sorted run at all. There are just delta and
image layers. However, when compaction starts, we look at the layers
that exist, and arrange them into levels, depending on their shapes.
That is ephemeral: when the compaction finishes, we forget that
information. This allows the new algorithm to work without any extra
bookkeeping. That makes it easier to transition from the old algorithm
to new, and back again.
There is just a new tenant config option to choose the compaction
algorithm. The default is "Legacy", meaning the current algorithm in
'main'. If you set it to "Tiered", the new algorithm is used.
3. A simulator, which implements the new abstraction.
The simulator can be used to analyze write and storage amplification,
without running a test with the full pageserver. It can also draw an SVG
animation of the simulation, to visualize how layers are created and
deleted.
To run the simulator:
cargo run --bin compaction-simulator run-suite
---------
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
This PR introduces a new vectored implementation of the read path.
The search is basically a DFS if you squint at it long enough.
LayerFringe tracks the next layers to visit and acts as our stack.
Vertices are tuples of (layer, keyspace, lsn range). Continuously
pop the top of the stack (most recent layer) and do all the reads
for one layer at once.
The search maintains a fringe (`LayerFringe`) which tracks all the
layers that intersect the current keyspace being searched. Continuously
pop the top of the fringe (layer with highest LSN) and get all the data
required from the layer in one go.
Said search is done on one timeline at a time. If data is still required for
some keys, then search the ancestor timeline.
Apart from the high level layer traversal, vectored variants have been
introduced for grabbing data from each layer type. They still suffer from
read amplification issues and that will be addressed in a different PR.
You might notice that in some places we duplicate the code for the
existing read path. All of that code will be removed when we switch
the non-vectored read path to proxy into the vectored read path.
In the meantime, we'll have to contend with the extra cruft for the sake
of testing and gentle releasing.
## Problem
One WAL record can actually produce an arbitrary amount of key value pairs.
This is problematic since it might cause our frozen layers to bloat past the
max allowed size of S3 single shot uploads.
[#6639](https://github.com/neondatabase/neon/pull/6639) introduced a "should roll"
check after every batch of `ingest_batch_size` (100 WAL records by default). This helps,
but the original problem still exists.
## Summary of changes
This patch moves the responsibility of rolling the currently open layer
to the `TimelineWriter`. Previously, this was done ad-hoc via calls
to `check_checkpoint_distance`. The advantages of this approach are:
* ability to split one batch over multiple open layers
* less layer map locking
* remove ad-hoc check_checkpoint_distance calls
More specifically, we track the current size of the open layer in the
writer. On each `put` check whether the current layer should be closed
and a new one opened. Keeping track of the currently open layer results
in less contention on the layer map lock. It only needs to be acquired
on the first write and on writes that require a roll afterwards.
Rolling the open layer can be triggered by:
1. The distance from the last LSN we rolled at. This bounds the amount
of WAL that the safekeepers need to store.
2. The size of the currently open layer.
3. The time since the last roll. It helps safekeepers to regard
pageserver as caught up and suspend activity.
Closes#6624
Refactor out layer accesses so that we can have easy access to resident
layers, which are needed for number of cases instead of layers for
eviction. Simplifies the heatmap building by only using Layers, not
RemoteTimelineClient.
Cc: #5331
This PR refactors the `blob_io` code away from using slices towards
taking owned buffers and return them after use.
Using owned buffers will eventually allow us to use io_uring for writes.
part of https://github.com/neondatabase/neon/issues/6663
Depends on https://github.com/neondatabase/tokio-epoll-uring/pull/43
The high level scheme is as follows:
- call writing functions with the `BoundedBuf`
- return the underlying `BoundedBuf::Buf` for potential reuse in the
caller
NB: Invoking `BoundedBuf::slice(..)` will return a slice that _includes
the uninitialized portion of `BoundedBuf`_.
I.e., the portion between `bytes_init()` and `bytes_total()`.
It's a safe API that actually permits access to uninitialized memory.
Not great.
Another wrinkle is that it panics if the range has length 0.
However, I don't want to switch away from the `BoundedBuf` API, since
it's what tokio-uring uses.
We can always weed this out later by replacing `BoundedBuf` with our own
type.
Created an issue so we don't forget:
https://github.com/neondatabase/tokio-epoll-uring/issues/46
This PR reverts
- https://github.com/neondatabase/neon/pull/6589
- https://github.com/neondatabase/neon/pull/6652
because there's a performance regression that's particularly visible at
high layer counts.
Most likely it's because the switch to RwLock inflates the
```
inner: heavier_once_cell::OnceCell<ResidentOrWantedEvicted>,
```
size from 48 to 88 bytes, which, by itself is almost a doubling of the
cache footprint, and probably the fact that it's now larger than a cache
line also doesn't help.
See this chat on the Neon discord for more context:
https://discord.com/channels/1176467419317940276/1204714372295958548/1205541184634617906
I'm reverting 6652 as well because it might also have perf implications,
and we're getting close to the next release. We should re-do its changes
after the next release, though.
cc @koivunej
cc @ivaxer
There is currently no cleanup done after a delta layer creation error,
so delta layers can accumulate. The problem gets worse as the operation
gets retried and delta layers accumulate on the disk. Therefore, delete
them from disk (if something has been written to disk).
This was very useful in debugging the bugs fixed in #6410 and #6502.
There's a lot more we could do. This only adds the printing to delta
layers, not image layers, for example, and it might be useful to print
details of more record types. But this is a good start.
## Problem
There's no efficient way of querying the layer map for a range.
## Summary of changes
Introduce a range query for the layer map (`LayerMap::range_search`).
There's two broad steps to it:
1. Find all coverage changes for layers that intersect the queried range
(see `LayerCoverage::range_overlaps`).
The slightly tricky part is dealing with the start of the range. We can
either be aligned with a layer or not and we need
to treat these cases differently.
2. Iterate over the coverage changes and collect the result. For this we
use a two pointer approach: the trailing pointer tracks the start of the
current range (current location in the key space) and the forward
pointer tracks the next coverage change.
Plugging the range search into the read path is deferred to a future PR.
## Performance
I adapted the layer map benchmarks on a local branch. Range searches are
between 2x and 2.5x slower than point searches. That's in line with what I
expected since we query thelayer map twice.
Since `Timeline::get` will proxy to `Timeline::get_vectored` we can
special case the one element layer map range search
at that point.
Follows #6123
Closes: https://github.com/neondatabase/neon/issues/5342
The approach here is to avoid using `Layer` from secondary tenants, and
instead make the eviction types (e.g. `EvictionCandidate`) have a
variant that carries a Layer for attached tenants, and a different
variant for secondary tenants.
Other changes:
- EvictionCandidate no longer carries a `Timeline`: this was only used
for providing a witness reference to remote timeline client.
- The types for returning eviction candidates are all in
disk_usage_eviction_task.rs now, whereas some of them were in
timeline.rs before.
- The EvictionCandidate type replaces LocalLayerInfoForDiskUsageEviction
type, which was basically the same thing.
## Problem
Previously, `GET /v1/tenant/:tenant_id/timeline` and `GET
/v1/tenant/:tenant_id/timeline/:timeline_id`
would bump the priority of the background task which computes the
initial logical size by cancelling
the wait on the synchronisation semaphore. However, the request would
still return an approximate
logical size. It's undesirable to force background work for a status
request.
## Summary of changes
This PR updates the priority used by the timeline status request such
that they don't do priority boosting
by default anymore. An optional query parameter,
`force-await-initial-logical-size`, is added for both
mentioned endpoints. When set to true, it will skip the concurrency
limiting semaphore and wait
for the background task to complete before returning the exact logical
size.
In order to exercise this behaviour in a test I had to add an extra
failpoint. If you think it's too intrusive,
it can be removed.
Also fixeda small bug where the cancellation of a download is reported as an
opaque download failure upstream. This caused `test_location_conf_churn`
to fail at teardown due to a WARN log line.
Closes https://github.com/neondatabase/neon/issues/6168
These functions don't need updating for sharding: it's fine for them to
remain shard-naive, as they're only used in the context of dumping a
layer file. The sharding metadata doesn't live in the layer file, it
lives in the index.
## Problem
Noticed while making other changes that there were `pub` items that were
unused.
## Summary of changes
- Make everything `pub(crate)` in metrics.rs, apart from items used from
`bin/`
- Fix the timelines eviction metric: it was never being incremented
- Remove an unused ephemeral_bytes counter.
## Problem
For context, this problem was observed in a research project where we
try to make neon run in multiple regions and I was asked by @hlinnaka to
make this PR.
In our project, we use the pageserver in a non-conventional way such
that we would send a larger number of requests to the pageserver than
normal (imagine postgres without the buffer pool). I measured the time
from the moment a WAL record left the safekeeper to when it reached the
pageserver
([code](e593db1f5a/pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs (L282-L287)))
and observed that when the number of get_page_at_lsn requests was high,
the wal receiving time increased significantly (see the left side of the
graphs below).
Upon further investigation, I found that the delay was caused by this
line
d2ca410919/pageserver/src/tenant/timeline.rs (L2348)
The `get_layer_for_write` method is called for every value during WAL
ingestion and it tries to acquire layers write lock every time, thus
this results in high contention when read lock is acquired more
frequently.
## Summary of changes
It is unnecessary to call `get_layer_for_write` repeatedly for all
values in a WAL message since they would end up in the same memory layer
anyway, so I created the batched versions of `InMemoryLayer::put_value`,
`InMemoryLayer ::put_tombstone`, `Timeline::put_value`, and
`Timeline::put_tombstone`, that acquire the locks once for a batch of
values.
Additionally, `DatadirModification` is changed to store multiple
versions of uncommitted values, and `WalIngest::ingest_record()` can now
ingest records without immediately committing them.
With these new APIs, the new ingestion loop can be changed to commit for
every `ingest_batch_size` records. The `ingest_batch_size` variable is
exposed as a config. If it is set to 1 then we get the same behavior
before this change. I found that setting this value to 100 seems to work
the best, and you can see its effect on the right side of the above
graphs.
---------
Co-authored-by: John Spray <john@neon.tech>
