Taking continuous profiles every 20 seconds is likely too expensive (in
dollar terms). Let's try 60-second profiles. We can now interrupt
running profiles via `?force=true`, so this should be fine.
## Problem
It's only possible to take one CPU profile at a time. With Grafana
continuous profiling, a (low-frequency) CPU profile will always be
running, making it hard to take an ad hoc CPU profile at the same time.
Resolves#10072.
## Summary of changes
Add a `force` parameter for `/profile/cpu` which will end and return an
already running CPU profile, starting a new one for the current caller.
## Problem
Unlike CPU profiles, the `/profile/heap` endpoint can't automatically
generate SVG flamegraphs. This requires the user to install and use
`pprof` tooling, which is unnecessary and annoying.
Resolves#10203.
## Summary of changes
Add `format=svg` for the `/profile/heap` route, and generate an SVG
flamegraph using the `inferno` crate, similarly to what `pprof-rs`
already does for CPU profiles.
## Problem
Jemalloc heap profiles aren't symbolized. This is inconvenient, and
doesn't work with Grafana Cloud Profiles.
Resolves#9964.
## Summary of changes
Symbolize the heap profiles in-process, and strip unnecessary cruft.
This uses about 100 MB additional memory to cache the DWARF information,
but I believe this is already the case with CPU profiles, which use the
same library for symbolization. With cached DWARF information, the
symbolization CPU overhead is negligible.
Example profiles:
*
[pageserver.pb.gz](https://github.com/user-attachments/files/18141395/pageserver.pb.gz)
*
[safekeeper.pb.gz](https://github.com/user-attachments/files/18141396/safekeeper.pb.gz)
## Problem
We don't have good observability for memory usage. This would be useful
e.g. to debug OOM incidents or optimize performance or resource usage.
We would also like to use continuous profiling with e.g. [Grafana Cloud
Profiles](https://grafana.com/products/cloud/profiles-for-continuous-profiling/)
(see https://github.com/neondatabase/cloud/issues/14888).
This PR is intended as a proof of concept, to try it out in staging and
drive further discussions about profiling more broadly.
Touches https://github.com/neondatabase/neon/issues/9534.
Touches https://github.com/neondatabase/cloud/issues/14888.
Depends on #9779.
Depends on #9780.
## Summary of changes
Adds a HTTP route `/profile/heap` that takes a heap profile and returns
it. Query parameters:
* `format`: output format (`jemalloc` or `pprof`; default `pprof`).
Unlike CPU profiles (see #9764), heap profiles are not symbolized and
require the original binary to translate addresses to function names. To
make this work with Grafana, we'll probably have to symbolize the
process server-side -- this is left as future work, as is other output
formats like SVG.
Heap profiles don't work on macOS due to limitations in jemalloc.
## Problem
We don't have a convenient way to gather CPU profiles from a running
binary, e.g. during production incidents or end-to-end benchmarks, nor
during microbenchmarks (particularly on macOS).
We would also like to have continuous profiling in production, likely
using [Grafana Cloud
Profiles](https://grafana.com/products/cloud/profiles-for-continuous-profiling/).
We may choose to use either eBPF profiles or pprof profiles for this
(pending testing and discussion with SREs), but pprof profiles appear
useful regardless for the reasons listed above. See
https://github.com/neondatabase/cloud/issues/14888.
This PR is intended as a proof of concept, to try it out in staging and
drive further discussions about profiling more broadly.
Touches #9534.
Touches https://github.com/neondatabase/cloud/issues/14888.
## Summary of changes
Adds a HTTP route `/profile/cpu` that takes a CPU profile and returns
it. Defaults to a 5-second pprof Protobuf profile for use with e.g.
`pprof` or Grafana Alloy, but can also emit an SVG flamegraph. Query
parameters:
* `format`: output format (`pprof` or `svg`)
* `frequency`: sampling frequency in microseconds (default 100)
* `seconds`: number of seconds to profile (default 5)
Also integrates pprof profiles into Criterion benchmarks, such that
flamegraph reports can be taken with `cargo bench ... --profile-duration
<seconds>`. Output under `target/criterion/*/profile/flamegraph.svg`.
Example profiles:
* pprof profile (use [`pprof`](https://github.com/google/pprof)):
[profile.pb.gz](https://github.com/user-attachments/files/17756788/profile.pb.gz)
* Web interface: `pprof -http :6060 profile.pb.gz`
* Interactive flamegraph:
[profile.svg.gz](https://github.com/user-attachments/files/17756782/profile.svg.gz)
## Problem
new clippy warnings on nightly.
## Summary of changes
broken up each commit by warning type.
1. Remove some unnecessary refs.
2. In edition 2024, inference will default to `!` and not `()`.
3. Clippy complains about doc comment indentation
4. Fix `Trait + ?Sized` where `Trait: Sized`.
5. diesel_derives triggering `non_local_defintions`
## Problem
Storage controller had basically no metrics.
## Summary of changes
1. Migrate the existing metrics to use Conrad's
[`measured`](https://docs.rs/measured/0.0.14/measured/) crate.
2. Add metrics for incoming http requests
3. Add metrics for outgoing http requests to the pageserver
4. Add metrics for outgoing pass through requests to the pageserver
5. Add metrics for database queries
Note that the metrics response for the attachment service does not use
chunked encoding like the rest of the metrics endpoints. Conrad has
kindly extended the crate such that it can now be done. Let's leave it
for a follow-up since the payload shouldn't be that big at this point.
Fixes https://github.com/neondatabase/neon/issues/6875
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.
* lower level on auth success from info to debug (fixes#5820)
* don't log stacktraces on auth errors (as requested on slack). we do this by introducing an `AuthError` type instead of using `anyhow` and `bail`.
* return errors that have been censored for improved security.
## Problem
For quickly rotating JWT secrets, we want to be able to reload the JWT
public key file in the pageserver, and also support multiple JWT keys.
See #4897.
## Summary of changes
* Allow directories for the `auth_validation_public_key_path` config
param instead of just files. for the safekeepers, all of their config options
also support multiple JWT keys.
* For the pageservers, make the JWT public keys easily globally swappable
by using the `arc-swap` crate.
* Add an endpoint to the pageserver, triggered by a POST to
`/v1/reload_auth_validation_keys`, that reloads the JWT public keys from
the pre-configured path (for security reasons, you cannot upload any
keys yourself).
Fixes#4897
---------
Co-authored-by: Heikki Linnakangas <heikki@neon.tech>
Co-authored-by: Joonas Koivunen <joonas@neon.tech>
This is a full switch, fs io operations are also tokio ones, working through
thread pool. Similar to pageserver, we have multiple runtimes for easier `top`
usage and isolation.
Notable points:
- Now that guts of safekeeper.rs are full of .await's, we need to be very
careful not to drop task at random point, leaving timeline in unclear
state. Currently the only writer is walreceiver and we don't have top
level cancellation there, so we are good. But to be safe probably we should
add a fuse panicking if task is being dropped while operation on a timeline
is in progress.
- Timeline lock is Tokio one now, as we do disk IO under it.
- Collecting metrics got a crutch: since prometheus Collector is
synchronous, it spawns a thread with current thread runtime collecting data.
- Anything involving closures becomes significantly more complicated, as
async fns are already kinda closures + 'async closures are unstable'.
- Main thread now tracks other main tasks, which got much easier.
- The only sync place left is initial data loading, as otherwise clippy
complains on timeline map lock being held across await points -- which is
not bad here as it happens only in single threaded runtime of main thread.
But having it sync doesn't hurt either.
I'm concerned about performance of thread pool io offloading, async traits and
many await points; but we can try and see how it goes.
fixes https://github.com/neondatabase/neon/issues/3036
fixes https://github.com/neondatabase/neon/issues/3966
We now spawn a new task for every HTTP request, and wait on the
JoinHandle. If Hyper drops the Future, the spawned task will keep
running. This protects the rest of the pageserver code from unexpected
async cancellations.
This creates a CancellationToken for each request and passes it to the
handler function. If the HTTP request is dropped by the client, the
CancellationToken is signaled. None of the handler functions make use
for the CancellationToken currently, but they now they could.
The CancellationToken arguments also work like documentation. When
you're looking at a function signature and you see that it takes a
CancellationToken as argument, it's a nice hint that the function might
run for a long time, and won't be async cancelled. The default
assumption in the pageserver is now that async functions are not
cancellation-safe anyway, unless explictly marked as such, but this is a
nice extra reminder.
Spawning a task for each request is OK from a performance point of view
because spawning is very cheap in Tokio, and none of our HTTP requests
are very performance critical anyway.
Fixes issue #3478
Previously, you used it like this:
|r| RequestSpan(my_handler).handle(r)
But I don't see the point of the RequestSpan struct. It's just a
wrapper around the handler function. With this commit, the call
becomes:
|r| request_span(r, my_handler)
Which seems a little simpler.
At first I thought that the RequestSpan struct would allow "chaining"
other kinds of decorators like RequestSpan, so that you could do
something like this:
|r| CheckPermissions(RequestSpan(my_handler)).handle(r)
But it doesn't work like that. If each of those structs wrap a handler
*function*, it would actually look like this:
|r| CheckPermissions(|r| RequestSpan(my_handler).handle(r))).handle(r)
This commit doesn't make that kind of chaining any easier, but seems a
little more straightforward anyway.
Require the error type to be ApiError. It implicitly required that
anyway, because the function used error::handler, which downcasted the
error to an ApiError. If the error was in fact anything else than
ApiError, it would just panic. Better to check it at compilation time.
Also make the last-resort error handler more forgiving, so that it
returns an 500 Internal Server error response, instead of panicking,
if a request handler returns some other error than an ApiError.
noticed while describing `RequestSpan`, this fix will omit the otherwise
logged message about request being cancelled when panicking in the
request handler. this was missed on #4064.
Add a simple disarmable dropguard to log if request is cancelled before
it is completed. We currently don't have this, and it makes for
difficult to know when the request was dropped.
Adds a newtype that creates a span with request_id from
https://github.com/neondatabase/neon/pull/3708 for every HTTP request
served.
Moves request logging and error handlers under the new wrapper, so every request-related event now is logged under the request span.
For compatibility reasons, error handler is left on the general router, since not every service uses the new handler wrappers yet.
## Describe your changes
## Issue ticket number and link
#3479
## Checklist before requesting a review
- [x] 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.
We started rather frequently scrap some apis for metadata. This includes
layer eviction tester, I believe console does that too.
It should eliminate these logs:
https://neonprod.grafana.net/goto/rr_ace1Vz?orgId=1 (Note the rate
around 2k messages per minute)
This patch adds a LaunchTimestamp type to the `metrics` crate,
along with a `libmetric_` Prometheus metric.
The initial user is pageserver.
In addition to exposing the Prometheus metric, it also reproduces
the launch timestamp as a header in the API responses.
The motivation for this is that we plan to scrape the pageserver's
/v1/tenant/:tenant_id/timeline/:timeline_id/layer
HTTP endpoint over time. It will soon expose access metrics (#3496)
which reset upon process restart. We will use the pageserver's launch
ID to identify a restart between two scrape points.
However, there are other potential uses. For example, we could use
the Prometheus metric to annotate Grafana plots whenever the launch
timestamp changes.
There will be different scopes for those two, so authorization code should be different.
The `check_permission` function is now not in the shared library. Its implementation
is very similar to the one which will be added for Safekeeper. In fact, we may reuse
the same existing root-like 'PageServerApi' scope, but I would prefer to have separate
root-like scopes for services.
Also, generate_management_token in tests is generate_pageserver_token now.
We had a problem where almost all of the threads were waiting on a futex syscall. More specifically:
- `/metrics` handler was inside `TimelineCollector::collect()`, waiting on a mutex for a single Timeline
- This exact timeline was inside `control_file::FileStorage::persist()`, waiting on a mutex for Lazy initialization of `PERSIST_CONTROL_FILE_SECONDS`
- `PERSIST_CONTROL_FILE_SECONDS: Lazy<Histogram>` was blocked on `prometheus::register`
- `prometheus::register` calls `DEFAULT_REGISTRY.write().register()` to take a write lock on Registry and add a new metric
- `DEFAULT_REGISTRY` lock was already taken inside `DEFAULT_REGISTRY.gather()`, which was called by `/metrics` handler to collect all metrics
This commit creates another Registry with a separate lock, to avoid deadlock in a case where `TimelineCollector` triggers registration of new metrics inside default registry.
- Enabled process exporter for storage services
- Changed zenith_proxy prefix to just proxy
- Removed old `monitoring` directory
- Removed common prefix for metrics, now our common metrics have `libmetrics_` prefix, for example `libmetrics_serve_metrics_count`
- Added `test_metrics_normal_work`
