Commit Graph

6 Commits

Author SHA1 Message Date
John Spray
73c6626b38 pageserver: stabilize & refine controller scale test (#8971)
## Problem

We were seeing timeouts on migrations in this test.

The test unfortunately tends to saturate local storage, which is shared
between the pageservers and the control plane database, which makes the
test kind of unrealistic. We will also want to increase the scale of
this test, so it's worth fixing that.

## Summary of changes

- Instead of randomly creating timelines at the same time as the other
background operations, explicitly identify a subset of tenant which will
have timelines, and create them at the start. This avoids pageservers
putting a lot of load on the test node during the main body of the test.
- Adjust the tenants created to create some number of 8 shard tenants
and the rest 1 shard tenants, instead of just creating a lot of 2 shard
tenants.
- Use archival_config to exercise tenant-mutating operations, instead of
using timeline creation for this.
- Adjust reconcile_until_idle calls to avoid waiting 5 seconds between
calls, which causes timelines with large shard count tenants.
- Fix a pageserver bug where calls to archival_config during activation
get 404
2024-10-15 09:31:18 +01:00
Folke Behrens
54d1185789 proxy: Unalias hyper1 and replace one use of hyper0 in test (#9324)
Leaves one final use of hyper0 in proxy for the health service,
which requires some coordinated effort with other services.
2024-10-09 12:44:17 +02:00
Conrad Ludgate
94a5ca2817 proxy: auth broker (#8855)
Opens http2 connection to local-proxy and forwards requests over with
all headers and body

closes https://github.com/neondatabase/cloud/issues/16039
2024-09-30 20:43:45 +01:00
Conrad Ludgate
12850dd5e9 proxy: remove dead code (#8847)
By marking everything possible as pub(crate), we find a few dead code
candidates.
2024-08-27 12:00:35 +01:00
Yuchen Liang
630cfbe420 refactor(pageserver): designated api error type for cancelled request (#7949)
Closes #7406.

## Problem

When a `get_lsn_by_timestamp` request is cancelled, an anyhow error is
exposed to handle that case, which verbosely logs the error. However, we
don't benefit from having the full backtrace provided by anyhow in this
case.

## Summary of changes

This PR introduces a new `ApiError` type to handle errors caused by
cancelled request more robustly.
-  A new enum variant `ApiError::Cancelled`
- Currently the cancelled request is mapped to status code 500.
- Need to handle this error in proxy's `http_util` as well.
- Added a failpoint test to simulate cancelled `get_lsn_by_timestamp`
request.

Signed-off-by: Yuchen Liang <yuchen@neon.tech>
2024-06-06 14:00:14 +00:00
Conrad Ludgate
c0ff4f18dc proxy: hyper1 for only proxy (#7073)
## Problem

hyper1 offers control over the HTTP connection that hyper0_14 does not.
We're blocked on switching all services to hyper1 because of how we use
tonic, but no reason we can't switch proxy over.

## Summary of changes

1. hyper0.14 -> hyper1
    1. self managed server
    2. Remove the `WithConnectionGuard` wrapper from `protocol2`
2. Remove TLS listener as it's no longer necessary
3. include first session ID in connection startup logs
2024-04-10 08:23:59 +00:00