pg_search is 46ish MB. All other remote extensions are around hundeds of
KB. 3 seconds is not long enough to download the tarball if the S3
gateway cache doesn't already contain a copy. According to our setup,
the cache is limited to 10 GB in size and anything that has not been
accessed for an hour is purged.
This is really bad for scaling to 0, even more so if you're the only
project actively using the extension in a production Kubernetes cluster.
Signed-off-by: Tristan Partin <tristan@neon.tech>
## Problem
Currently the following line below uses array subscript notation which
is confusing since `reqlsns` is not an array but just a pointer to a
struct.
```
XLogWaitForReplayOf(reqlsns[0].request_lsn);
```
## Summary of changes
Switch from array subscript notation to arrow operator to improve
readability of code.
Close#10620.
## Problem
Incorrect manipulations with iteration index in `lfc_cache_containsv`
## Summary of changes
```
- int this_chunk = Min(nblocks, BLOCKS_PER_CHUNK - chunk_offs);
+ int this_chunk = Min(nblocks - i, BLOCKS_PER_CHUNK - chunk_offs); int this_chunk = ```
- if (i + 1 >= nblocks)
+ if (i >= nblocks)
```
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
## Problem
USE_ASSERT _CHECKING is defined as empty entity. but it is checked using
#if
## Summary of changes
Replace `#if USE_ASSERT _CHECKING` with `#ifdef USE_ASSERT _CHECKING` as
done in other places in Postgres
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
Drop logical replication subscribers
before compute starts on a non-main branch.
Add new compute_ctl spec flag: drop_subscriptions_before_start
If it is set, drop all the subscriptions from the compute node
before it starts.
To avoid race on compute start, use new GUC
neon.disable_logical_replication_subscribers
to temporarily disable logical replication workers until we drop the
subscriptions.
Ensure that we drop subscriptions exactly once when endpoint starts on a
new branch.
It is essential, because otherwise, we may drop not only inherited, but
newly created subscriptions.
We cannot rely only on spec.drop_subscriptions_before_start flag,
because if for some reason compute restarts inside VM,
it will start again with the same spec and flag value.
To handle this, we save the fact of the operation in the database
in the neon.drop_subscriptions_done table.
If the table does not exist, we assume that the operation was never
performed, so we must do it.
If table exists, we check if the operation was performed on the current
timeline.
fixes: https://github.com/neondatabase/neon/issues/8790
Otherwise we might hit ERRORs in otherwise safe situations (such as user
queries), which isn't a great user experience.
## Problem
https://github.com/neondatabase/neon/pull/10376
## Summary of changes
Instead of accepting internal errors as acceptable, we ensure we don't
exceed our allocated usage.
# Refs
- fixes https://github.com/neondatabase/neon/issues/10309
- fixup of batching design, first introduced in
https://github.com/neondatabase/neon/pull/9851
- refinement of https://github.com/neondatabase/neon/pull/8339
# Problem
`Tenant::shutdown` was occasionally taking many minutes (sometimes up to
20) in staging and prod if the
`page_service_pipelining.mode="concurrent-futures"` is enabled.
# Symptoms
The issue happens during shard migration between pageservers.
There is page_service unavailability and hence effectively downtime for
customers in the following case:
1. The source (state `AttachedStale`) gets stuck in `Tenant::shutdown`,
waiting for the gate to close.
2. Cplane/Storcon decides to transition the target `AttachedMulti` to
`AttachedSingle`.
3. That transition comes with a bump of the generation number, causing
the `PUT .../location_config` endpoint to do a full `Tenant::shutdown` /
`Tenant::attach` cycle for the target location.
4. That `Tenant::shutdown` on the target gets stuck, waiting for the
gate to close.
5. Eventually the gate closes (`close completed`), correlating with a
`page_service` connection handler logging that it's exiting because of a
network error (`Connection reset by peer` or `Broken pipe`).
While in (4):
- `Tenant::shutdown` is stuck waiting for all `Timeline::shutdown` calls
to complete.
So, really, this is a `Timeline::shutdown` bug.
- retries from Cplane/Storcon to complete above state transitions, fail
with errors related to the tenant mgr slot being in state
`TenantSlot::InProgress`, the tenant state being
`TenantState::Stopping`, and the timelines being in
`TimelineState::Stopping`, and the `Timeline::cancel` being cancelled.
- Existing (and/or new?) page_service connections log errors `error
reading relation or page version: Not found: Timed out waiting 30s for
tenant active state. Latest state: None`
# Root-Cause
After a lengthy investigation ([internal
write-up](https://www.notion.so/neondatabase/2025-01-09-batching-deadlock-Slow-Log-Analysis-in-Staging-176f189e00478050bc21c1a072157ca4?pvs=4))
I arrived at the following root cause.
The `spsc_fold` channel (`batch_tx`/`batch_rx`) that connects the
Batcher and Executor stages of the pipelined mode was storing a `Handle`
and thus `GateGuard` of the Timeline that was not shutting down.
The design assumption with pipelining was that this would always be a
short transient state.
However, that was incorrect: the Executor was stuck on writing/flushing
an earlier response into the connection to the client, i.e., socket
write being slow because of TCP backpressure.
The probable scenario of how we end up in that case:
1. Compute backend process sends a continuous stream of getpage prefetch
requests into the connection, but never reads the responses (why this
happens: see Appendix section).
2. Batch N is processed by Batcher and Executor, up to the point where
Executor starts flushing the response.
3. Batch N+1 is procssed by Batcher and queued in the `spsc_fold`.
4. Executor is still waiting for batch N flush to finish.
5. Batcher eventually hits the `TimeoutReader` error (10min).
From here on it waits on the
`spsc_fold.send(Err(QueryError(TimeoutReader_error)))`
which will never finish because the batch already inside the `spsc_fold`
is not
being read by the Executor, because the Executor is still stuck in the
flush.
(This state is not observable at our default `info` log level)
6. Eventually, Compute backend process is killed (`close()` on the
socket) or Compute as a whole gets killed (probably no clean TCP
shutdown happening in that case).
7. Eventually, Pageserver TCP stack learns about (6) through RST packets
and the Executor's flush() call fails with an error.
8. The Executor exits, dropping `cancel_batcher` and its end of the
spsc_fold.
This wakes Batcher, causing the `spsc_fold.send` to fail.
Batcher exits.
The pipeline shuts down as intended.
We return from `process_query` and log the `Connection reset by peer` or
`Broken pipe` error.
The following diagram visualizes the wait-for graph at (5)
```mermaid
flowchart TD
Batcher --spsc_fold.send(TimeoutReader_error)--> Executor
Executor --flush batch N responses--> socket.write_end
socket.write_end --wait for TCP window to move forward--> Compute
```
# Analysis
By holding the GateGuard inside the `spsc_fold` open, the pipelining
implementation
violated the principle established in
(https://github.com/neondatabase/neon/pull/8339).
That is, that `Handle`s must only be held across an await point if that
await point
is sensitive to the `<Handle as Deref<Target=Timeline>>::cancel` token.
In this case, we were holding the Handle inside the `spsc_fold` while
awaiting the
`pgb_writer.flush()` future.
One may jump to the conclusion that we should simply peek into the
spsc_fold to get
that Timeline cancel token and be sensitive to it during flush, then.
But that violates another principle of the design from
https://github.com/neondatabase/neon/pull/8339.
That is, that the page_service connection lifecycle and the Timeline
lifecycles must be completely decoupled.
Tt must be possible to shut down one shard without shutting down the
page_service connection, because on that single connection we might be
serving other shards attached to this pageserver.
(The current compute client opens separate connections per shard, but,
there are plans to change that.)
# Solution
This PR adds a `handle::WeakHandle` struct that does _not_ hold the
timeline gate open.
It must be `upgrade()`d to get a `handle::Handle`.
That `handle::Handle` _does_ hold the timeline gate open.
The batch queued inside the `spsc_fold` only holds a `WeakHandle`.
We only upgrade it while calling into the various `handle_` methods,
i.e., while interacting with the `Timeline` via `<Handle as
Deref<Target=Timeline>>`.
All that code has always been required to be (and is!) sensitive to
`Timeline::cancel`, and therefore we're guaranteed to bail from it
quickly when `Timeline::shutdown` starts.
We will drop the `Handle` immediately, before we start
`pgb_writer.flush()`ing the responses.
Thereby letting go of our hold on the `GateGuard`, allowing the timeline
shutdown to complete while the page_service handler remains intact.
# Code Changes
* Reproducer & Regression Test
* Developed and proven to reproduce the issue in
https://github.com/neondatabase/neon/pull/10399
* Add a `Test` message to the pagestream protocol (`cfg(feature =
"testing")`).
* Drive-by minimal improvement to the parsing code, we now have a
`PagestreamFeMessageTag`.
* Refactor `pageserver/client` to allow sending and receiving
`page_service` requests independently.
* Add a Rust helper binary to produce situation (4) from above
* Rationale: (4) and (5) are the same bug class, we're holding a gate
open while `flush()`ing.
* Add a Python regression test that uses the helper binary to
demonstrate the problem.
* Fix
* Introduce and use `WeakHandle` as explained earlier.
* Replace the `shut_down` atomic with two enum states for `HandleInner`,
wrapped in a `Mutex`.
* To make `WeakHandle::upgrade()` and `Handle::downgrade()`
cache-efficient:
* Wrap the `Types::Timeline` in an `Arc`
* Wrap the `GateGuard` in an `Arc`
* The separate `Arc`s enable uncontended cloning of the timeline
reference in `upgrade()` and `downgrade()`.
If instead we were `Arc<Timeline>::clone`, different connection handlers
would be hitting the same cache line on every upgrade()/downgrade(),
causing contention.
* Please read the udpated module-level comment in `mod handle`
module-level comment for details.
# Testing & Performance
The reproducer test that failed before the changes now passes, and
obviously other tests are passing as well.
We'll do more testing in staging, where the issue happens every ~4h if
chaos migrations are enabled in storcon.
Existing perf testing will be sufficient, no perf degradation is
expected.
It's a few more alloctations due to the added Arc's, but, they're low
frequency.
# Appendix: Why Compute Sometimes Doesn't Read Responses
Remember, the whole problem surfaced because flush() was slow because
Compute was not reading responses. Why is that?
In short, the way the compute works, it only advances the page_service
protocol processing when it has an interest in data, i.e., when the
pagestore smgr is called to return pages.
Thus, if compute issues a bunch of requests as part of prefetch but then
it turns out it can service the query without reading those pages, it
may very well happen that these messages stay in the TCP until the next
smgr read happens, either in that session, or possibly in another
session.
If there’s too many unread responses in the TCP, the pageserver kernel
is going to backpressure into userspace, resulting in our stuck flush().
All of this stems from the way vanilla Postgres does prefetching and
"async IO":
it issues `fadvise()` to make the kernel do the IO in the background,
buffering results in the kernel page cache.
It then consumes the results through synchronous `read()` system calls,
which hopefully will be fast because of the `fadvise()`.
If it turns out that some / all of the prefetch results are not needed,
Postgres will not be issuing those `read()` system calls.
The kernel will eventually react to that by reusing page cache pages
that hold completed prefetched data.
Uncompleted prefetch requests may or may not be processed -- it's up to
the kernel.
In Neon, the smgr + Pageserver together take on the role of the kernel
in above paragraphs.
In the current implementation, all prefetches are sent as GetPage
requests to Pageserver.
The responses are only processed in the places where vanilla Postgres
would do the synchronous `read()` system call.
If we never get to that, the responses are queued inside the TCP
connection, which, once buffers run full, will backpressure into
Pageserver's sending code, i.e., the `pgb_writer.flush()` that was the
root cause of the problems we're fixing in this PR.
This reduces pressure on the OS TCP read buffer by increasing the
moments we read data out of the receive buffer, and increasing the
number of bytes we can pull from that buffer when we do reads.
## Problem
A backend may not always consume its prefetch data quick enough
## Summary of changes
We add a new function `prefetch_pump_state` which pulls as many prefetch
requests from the OS TCP receive buffer as possible, but without
blocking.
This thus reduces pressure on OS-level TCP buffers, thus increasing
throughput by limiting throttling caused by full TCP buffers.
## Problem
We have several serious data corruption incidents caused by mismatch of
get-age requests:
https://neondb.slack.com/archives/C07FJS4QF7V/p1723032720164359
We hope that the problem is fixed now. But it is better to prevent such
kind of problems in future.
Part of https://github.com/neondatabase/cloud/issues/16472
## Summary of changes
This PR introduce new V3 version of compute<->pageserver protocol,
adding tag to getpage response.
So now compute is able to check if it really gets response to the
requested page.
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
Apparently, we failed to do this bookkeeping in quite a few places...
## Problem
Fixes https://github.com/neondatabase/cloud/issues/22364
## Summary of changes
Add accounting of dropped requests. Note that this includes prefetches
dropped due to things like "PS connection dropped unexpectedly" or
"prefetch queue is already full", but *not* (yet?) "dropped due to
backend shutdown".
Using `min(0, ...)` causes us to fail to wait in most situations, so a
lack of data would be a hot wait loop, which is bad.
## Problem
We noticed high CPU usage in some situations
## Problem
When entry was dropped and password wasn't set, new entry
had uninitialized memory in controlplane adapter
Resolves: https://github.com/neondatabase/cloud/issues/14914
## Summary of changes
Initialize password in all cases, add tests.
Minor formatting for less indentation
Improved comments will help others when they read the code, and the log
messages will help others understand why the logical replication monitor
works the way it does.
Signed-off-by: Tristan Partin <tristan@neon.tech>
## Problem
LFC used_pages statistic is not updated in case of LFC resize (shrinking
`neon.file_cache_size_limit`)
## Summary of changes
Update `lfc_ctl->used_pages` in `lfc_change_limit_hook`
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
We have metrics for GetPage request latencies, but this is an extra
measure to capture requests that take way too long in the logs. The log
message is printed every 10 s, until the response is received:
```
PG:2024-12-09 16:02:07.715 GMT [1782845] LOG: [NEON_SMGR] [shard 0] no response received from pageserver for 10.000 s, still waiting (sent 10613 requests, received 10612 responses)
PG:2024-12-09 16:02:17.723 GMT [1782845] LOG: [NEON_SMGR] [shard 0] no response received from pageserver for 20.008 s, still waiting (sent 10613 requests, received 10612 responses)
PG:2024-12-09 16:02:19.719 GMT [1782845] LOG: [NEON_SMGR] [shard 0] received response from pageserver after 22.006 s
```
## Problem
See https://github.com/neondatabase/neon/issues/9961
Current implementation of prefetch buffer resize doesn't correctly
handle in-flight requests
## Summary of changes
1. Fix index of entry we should wait for if new prefetch buffer size is
smaller than number of in-flight requests.
2. Correctly set flush position
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
If the pageserver connection is lost while receiving the prefetch
request, the prefetch queue is cleared. The error message prints the
values from the prefetch slot, but because the slot was already cleared,
they're all zeros:
LOG: [NEON_SMGR] [shard 0] No response from reading prefetch entry 0:
0/0/0.0 block 0. This can be caused by a concurrent disconnect
To fix, make local copies of the values.
In the passing, also add a sanity check that if the receive() call
succeeds, the prefetch slot is still intact.
## Problem
See https://neondb.slack.com/archives/C04DGM6SMTM/p1732110190129479
We observe the following error in the logs
```
[XX000] ERROR: [NEON_SMGR] [shard 3] Incorrect prefetch read: status=1 response=0x7fafef335138 my=128 receive=128
```
most likely caused by changing `neon.readahead_buffer_size`
## Summary of changes
1. Copy shard state
2. Do not use prefetch_set_unused in readahead_buffer_resize
3. Change prefetch buffer overflow criteria
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
In ea32f1d0a3, Matthias added a feature to
our extension to expose more granular wait events. However, due to the
typo, those wait events were never registered, so we used the more
generic wait events instead.
Signed-off-by: Tristan Partin <tristan@neon.tech>
This GUC will drop replication slots if the size of the
pg_logical/snapshots directory (not including temp snapshot files)
becomes larger than the specified size. Keeping the size of this
directory smaller will help with basebackup size from the pageserver.
Part-of: https://github.com/neondatabase/neon/issues/8619
Signed-off-by: Tristan Partin <tristan@neon.tech>
Fix direct reading from WAL buffers.
Pointer wasn't advanced which resulted in sending corrupted WAL if part
of read used WAL buffers and part read from the file. Also move it to
neon_walreader so that e.g. replication could also make use of it.
ref https://github.com/neondatabase/cloud/issues/19567
We will only have a C string if the specified role is a string.
Otherwise, we need to resolve references to public, current_role,
current_user, and session_user.
Fixes: https://github.com/neondatabase/cloud/issues/19323
Signed-off-by: Tristan Partin <tristan@neon.tech>
As a DBaaS provider, Neon needs to provide a stable platform for
customers to build applications upon. At the same time however, we also
need to enable customers to use the latest and greatest technology, so
they can prototype their work, and we can solicit feedback. If all
extensions are treated the same in terms of stability, it is hard to
meet that goal.
There are now two new GUCs created by the Neon extension:
neon.allow_unstable_extensions: This is a session GUC which allows
a session to install and load unstable extensions.
neon.unstable_extensions: This is a comma-separated list of extension
names. We can check if a CREATE EXTENSION statement is attempting to
install an unstable extension, and if so, deny the request if
neon.allow_unstable_extensions is not set to true.
Signed-off-by: Tristan Partin <tristan@neon.tech>
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
neon.c is getting crowded and the logical replication slot monitor is
a good candidate for reorganization. It is very self-contained, and
being in a separate file will make it that much easier to find.
Signed-off-by: Tristan Partin <tristan@neon.tech>
A sizeof on a pointer on a 64 bit machine is 8 bytes whereas
Entry::old_name is a 64 byte array of characters. There was most likely
no fallout since the string would start with NUL bytes, but best to fix
nonetheless.
Signed-off-by: Tristan Partin <tristan@neon.tech>
At least as far as removing individual files goes, this is the best
pattern for removing. I can't say the same for removing directories, but
I went ahead and changed those to `$(RM) -r` anyway.
Signed-off-by: Tristan Partin <tristan@neon.tech>
## Problem
See
https://neondb.slack.com/archives/C033A2WE6BZ/p1729007738526309?thread_ts=1722942856.987979&cid=C033A2WE6BZ
When replica receives WAL record which target page is not present in
shared buffer, we evict this page from LFC.
If all pages from the LFC chunk are evicted, then chunk is moved to the
beginning of LRU least to force it reuse.
Unfortunately access_count is not checked and if the entry is access at
this moment then this operation can cause LRU list corruption.
## Summary of changes
Check `access_count` in `lfc_evict`
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
## Problem
There is double update of resize cache in `put_rel_truncation`
Also `page_server_request` contains check that fork is MAIN_FORKNUM
which
1. is incorrect (because Vm/FSM pages are shreded in the same way as
MAIN fork pages and
2. is redundant because `page_server_request` is never called for `get
page` request so first part to OR condition is always true.
## Summary of changes
Remove redundant code
## Checklist before requesting a review
- [ ] I have performed a self-review of my code.
- [ ] If it is a core feature, I have added thorough tests.
- [ ] Do we need to implement analytics? if so did you add the relevant
metrics to the dashboard?
- [ ] If this PR requires public announcement, mark it with
/release-notes label and add several sentences in this section.
## Checklist before merging
- [ ] Do not forget to reformat commit message to not include the above
checklist
---------
Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>
We now also track:
- Number of PS IOs in-flight
- Number of pages cached by smgr prefetch implementation
- IO timing histograms for LFC reads and writes, per IO issued
## Problem
There's little insight into the timing metrics of LFC, and what the
prefetch state of each backend is.
This changes that, by measuring (and subsequently exposing) these data
points.
## Summary of changes
- Extract IOHistogram as separate type, rather than a collection of
fields on NeonMetrics
- others, see items above.
Part of https://github.com/neondatabase/neon/issues/8926
## Problem
On macOS:
```
/Users/runner/work/neon/neon//pgxn/neon/file_cache.c:623:19: error: variable 'has_remaining_pages' is used uninitialized whenever 'for' loop exits because its condition is false [-Werror,-Wsometimes-uninitialized]
```
## Summary of changes
- Initialise `has_remaining_pages` with `false`
It's not a bug, the variable is initialized when it's used, but the
compiler isn't smart enough to see that through all the conditions.
Part of the cleanup issue #9217.
/pgxn/neon/walproposer_compat.c:192:9: warning: function ‘WalProposerLibLog’ might be a candidate for ‘gnu_printf’ format attribute [-Wsuggest-attribute=format]
192 | vsnprintf(buf, sizeof(buf), fmt, args);
| ^~~~~~~~~
The warning:
warning: ISO C90 forbids mixed declarations and code [-Wdeclaration-after-statement]
It's PostgreSQL project style to stick to the old C90 style.
(Alternatively, we could disable it for our extension.)
Part of the cleanup issue #9217.
Prototypes for neon_writev(), neon_readv(), and neon_regisersync()
were missing. But instead of adding the missing prototypes, mark all
the smgr functions 'static'.
Part of the cleanup issue #9217.
