Merge upstream 0.5.0 #9
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
It seems like field->perm might be uninitialized memory depending on how it is allocated. I ran into an issue where different sockets had different file permissions, and some of those sockets weren't readable by the user which created it. This behavior probably started in https://github.com/twitter/twemproxy/pull/311/files
Fix indented bulleted list and add monospaced formatting to command flags within paragraphs
Add Docker support
Fix README typo
This detects some inconsistent uses of indentation and inconsistent space after casts, but has a lot more false positives. https://clang.llvm.org/docs/ClangFormat.html#script-for-patch-reformatting may be of use for saving time for contributors reformatting new patches
Plugins for editors/IDEs exist that can enforce some of these rules
Add editorconfig matching the c style guide, fix typos in c style guide
Running it in docker makes it easier to run this in other CI providers and to easily reproduce any failures seen in Travis locally. Move to python 3
This is useful if the tests hang due to a change. Otherwise, log capturing would prevent you from seeing the logs.
Run in multiple redis versions Update some unit tests and reorganize docker build steps
Fix failing unit tests, run in Docker
If hostnames are used instead of ip addresses for all hosts within a pool, and dns lookup fails, then get/multiget will hang indefinitely **instead of** responding with an error such as `SERVER_ERROR Host is down`. (I expect a client would detect this and close the connection, but this is not ideal, the client timeout could be a second) - Both redis and memcache protocols are affected - If a server resolves but is down, then `get` does respond with `SERVER_ERROR Host is down`. I suspect that's because memcached get is implemented to coalesce responses from multiple backend servers, even when there's only one key, and this is likely a bug specific to handling coalescing when there's no attempt to send the request to a backend server Because connection attempts are async but dns lookup is asynchronous, there's a non-null server connection for a host that's unavailable but a null server connection for a host that has a dns lookup error, and these end up using different code paths. (e.g. the former will call server_close() which does adjust nfrag_sent) Fixes twitter#596
Travis CI builds have been queueing for hours for some builds. This a simple docker script.
…ions Switch to GitHub actions for builds
…erance Don't hang for dns error responding to fragmented requests
Add a clang-format file approximating the c style guide
ncontinuum is `160 * the number of **non-ejected** hosts` for the ketama distribution, where 160 is the ketama consistent hashing points per server. However, the index returned by msg_backend_idx is a number between 0..nservers-1. - This is a waste of time zeroing out 160 times as much memory as needed - When there are 0 servers, this previously malloc()ed a pointer of size 0 bytes and nutcracker would write to invalid memory. Also, update redis nosetest expectation because newer redis versions changed the error message to mention replica instead. Fixes twitter#563 for redis and memcached. See the discussion there for more details.
Fragmentation would likely be slower. It would involve splitting up the request into multiple smaller requests (allocating and zeroing out an array) and concatenating strings to rebuild the original request. Fragmentation is also more likely to be prone to edge cases such as those mentioned in twitter#597 and twitter#595
…-of-ncontinuum Use nservers instead of ncontinuum to fix segfault when all hosts are ejected
Always initialize file permissions field for unix domain socket
…len and make sure hostlen is equal or less than KETAMA_MAX_HOSTLEN Amended version of twitter#567 Amended to remove the trailing null byte when logging if this does truncate, which shouldn't happen in practice Co-Authored-By: Tyson Andre <[email protected]> Co-Authored-By: 李广博 <[email protected]>
Fix possible crash copying malformed memcache value responses
Make error messages starting the stats server clearer
The last ubuntu release that this PPA supported was from 2016 (ubuntu 16.10). If the ppa receives updates, this can be reverted. Closes twitter#625 Closes twitter#588
Update documentation on running tests, config options, etc.
Update unit test for redis/memcache request testing
Warn about `strdup` being missing during `configure` for libyaml. (I was checking compatibility with older gcc and should have used `-std=gnu99`, not `-std=c99`) Avoid warning about pointer type casts and fallthrough. Add tests of remaining supported hash functions being compatible with libmemcached results.
Add miscellaneous tests, fix compilation warnings
See ChangeLog for details. There have been a lot of changes since 0.4.1 was published in 2015. I've tested many of these on a fork in combination with the heartbeat/failover/sentinel patches (planned for 0.6.0) and expect them to be a significant improvement over 0.4.1, but there may be build/runtime/packaging issues I'm not aware of on alternate configurations or OSes. twemproxy 0.5.0 will use github releases for distribution tarballs
Release twemproxy 0.5.0-RC1
Release twemproxy 0.5.0
zeriyoshi
force-pushed
the
merge_upstream_0.5.0
branch
from
3e7553a to
bf754e2
Compare
|
upstream で入った機能 ( |
|
完全に当時の記憶がなくなっているのと、upstream も結構進んでいるようなのでやりなおしたほうが早そうです |
|
rebase するのではなく upstream にこちら側で入れた修正を cherry-pick した方が早いかも |
|
@taka-oyama 久しぶりです😀 一つ聞きたいのですが、こちらの twemproxy は、今もメンテしていますか？ |
|
してますが、新しい機能とかはあまり考えてない状態です。 |
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.
No description provided.