[pull] main from cloudflare:main #1
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
![pull[bot]](https://avatars.githubusercontent.com/in/12910?s=60&v=4){kind=small}
This change allows customized HTTP1 reason phrase to be used and proxied.
Includes-commit: 069a792 Replicated-from: #209 Signed-off-by: Benjamin Leggett <[email protected]>
Allow generating custom responses to purge requests (requests to invalidate or delete from the HTTP cache).
SO_RCVBUF needs to be set before connect otherwise tcp settings will not be finalized correctly. And since autotuning is not engaged in this case, the throughput will suffer.
Provide default and non_exhaustive for TcpSocketOptions to future proof its changes. Also fix an issue that TcpScoketOptions can't be used for IPv4 sockets at all.
There is still a bug but change to this to help debug.
It is possible that when two threads try to release streams of the same h2 connection, one thread would register the connection in the in_use_pool while the other register it in the idle_pool. The solution is to put a lock so that the connection can only appear in one pool at a time.
Make it async and more natural.
These APIs show how much time a socket spends on waiting for data and the time waiting for data to be sent. They are good indicators of performance bottlenecks.
These functions allow to use stale values while performing updates in the background to minimize the lookup latency.
poll_capacity won't return if 0 capacity is requested.
This moves the call to `seek` on a file's cursor from the `HttpProxy` state machine to the `ServeFromCache` state machine. If it is able to seek, then the range is set to `None` so that the range filter is deactivated.
calling std::process::exit during shutdown. --- Fix pingora-core Server formatting to pass ci check. --- Add doc comments for ShutdownSignal --- Remove unused mut self from run_forever --- Add Debug derive for Shutdown signal --- Add docs for shutdown handling Includes-commit: 7f5c988 Includes-commit: b7ffdc5 Includes-commit: 0096b01 Includes-commit: 02c1ea1 Includes-commit: 05e0907 Includes-commit: 11881c5 Replicated-from: #493
Those names was created when the downstream only supported h1, which is no longer the case.
Since this change response_duplex(_vec) will return true when the body write finishes even if the end_stream flag is not set. This change makes pingora-proxy stop polling from the downstream connection when the response finishes already. Before this change, in a rare case where the h2 upstream responds the complete body according to its content-length header but end_stream is not received yet, the h1 downstream would already consider sending the next request over the reused connection because h1 doesn't rely on any explicit end_stream signal when content-length is used. In this case, the next request would cause an error.
Since the previous change that signals the downstream to finish when body write finishes, the downstream pipe would close before the upstream sends its final finishing HttpTask. This is not an error.
…ks succeed Co-authored-by: Viacheslav Biriukov <[email protected]
Also fix an issue where a new uri using set_uri may not be used if the raw_path_fallback was also in use.
Prior to this, invalidations did not affect the eviction system so the amount of available storage could be underutilized.
This allows the hit handler to tell the eviction system a weight to initialize the asset with if it is not already registered.
Dangling cache locks could occur when there were simply many readers at once waiting for the lock, the reason being the available permits could immediately be used up upon unlocking and before the WritePermit is dropped, resulting in a false positive dangling lock. Instead, just mark the WritePermit as finished when the cache key lock is unlocked as a result of having the permit. The most intuitive way to do this is to "turn in" the WritePermit back to the CacheKeyLock that created it when unlocking, so that the CacheKeyLock can mark it as finished.
The write permit a subrequest is holding onto may be dropped if the subrequest encounters an error. When this occurs, unlock the permit as a transient error.
Includes-commit: 256a939 Replicated-from: #484 Signed-off-by: shyunny <[email protected]>
Allow session reuse if configured in `fail_to_proxy`. Applies only to errors prior to proxying upstream, for now. Previously any short-circuiting error during request processing would result in closing the connection to downstream. This is unnecessary if there was an internal error that simply resulted in an error response being written downstream. Note that any `set_keepalive` on the Session itself is still respected, and this API currently does not overwrite reuse decisions made while proxying upstream. The user is responsible for implications of reusing the session in this manner, such as draining any remaining request body.
Delay it to later. This should help boosting how fast the server can accept now connections.
Rustls only supports v4
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot]
Can you help keep this open source service alive? 💖 Please sponsor : )