Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Couchbase 26.2 #2

Merged
merged 10,000 commits into from
Jan 25, 2024
Merged

Couchbase 26.2 #2

merged 10,000 commits into from
Jan 25, 2024

Conversation

stevewatanabe
Copy link
Member

This incorporates the following Couchbase customization

MB-56037: Fix incompatibility with erlang 24.3.4.4 TLS client

See https://github.com/erlang/otp/issues/6772
for more info

RaimoNiskanen and others added 30 commits October 18, 2023 12:15
Add in cacert directories for modern illumos distributions to fix no_cacerts_found error.
…P-18814' into maint

* myself/dgud/public_key/cacerts-sunos/OTP-18814:
  Add OmniOS (illumos) cacert directories
Reduces the time-out message passing during high load, since
the server no longer acts as a proxy for client time-out messages.

There is still one timer per request towards the port driver,
instead of one per requesting client, and the clients use
receive...after.
For some reason the old Erlang Public License had gotten into some files.
This commit removes the unintended EPL license header in all files where it
appears and replace it with the intended Apache 2.0 license header.
wsl.exe is not available on x32 cmd since it is in the the 64b path.
Make the select and completion tag types more expressive.

OTP-18818
* lukas/gh/fix-prebuilt-cache-handling:
  Update GitHub Action Versions
  gh: NO_CACHE if too many deleted
  gh: Pass args as env vars as DELETED can become very large
  gh: Rebuild bootstrap if any bootstrap apps changed
  gh: Fix argument bug when changing what has changed in prebuilt
  gh: Build correct docker image for cross, 32 and clang
  gh: Split otp_src and otp_cache into separate cache actions
  gh: Pin kerl version
  gh: Fix perf install
  gh: Don't use key for cache restore
  gh: Backport github actions to OTP-26
Fix so that nested tuples doesn't crash the autocompletion
Fix so that the autocompletion doesn't suggest closing parenthesis
  since this often times interfere with completion of module names
  when the expected parameter type is atom in a function.
Fix so that functions with almost the same name or differenct cases
  doesn't get removed from autocompletion suggestions.
* dgud/ssl/fix-test-win-x32:
  Fix finding openssl on windows x32
A test run caused the (mib-suite) traffic test case
to enter an infinit loop. A syntax error was detected
(*not* part of the test case) and this resultet
in the test code enter a infinit loop (sending
syntax errors between (one of the) mg and mgc).
Attempted to brake out of this by cancel and then
disconnect.

A following test case failure, in another test suite,
because the nodes, with apps and config, where left running
after the failure. Causing the next test case to fail.
rickard-green and others added 28 commits December 9, 2023 16:59
…/rm_unused_makewhatis/25/OTP-18899

* rickard/rm_unused_makewhatis/24/OTP-18899:
  [erts] Remove unused makewhatis script
…fix/25/OTP-18896

* rickard/license-fix/24/OTP-18896:
  Fixes of license and copyright information
…fix/26/OTP-18896

* rickard/license-fix/25/OTP-18896:
  Update license information for OTP 25
  Fixes of license and copyright information
* rickard/rm_unused_makewhatis/25/OTP-18899:
  [erts] Remove unused makewhatis script
* rickard/license-fix/26/OTP-18896:
  Update license information for OTP 26
  Update license information for OTP 25
  Fixes of license and copyright information
* rickard/md5/26/OTP-18877:
  Update COPYRIGHT
…ted-words

remove repeated words and small typos
commit 91a63f5
Author:     Anders Nygren <anders.nygren@gmail.com>
AuthorDate: Thu Jun 30 15:58:25 2016 -0500
Commit:     Anders Nygren <anders.nygren@gmail.com>
CommitDate: Mon Aug 8 11:00:25 2016 -0500

    MB-19708: Change restart values for os_mon

    Under heavy CPU oad on windows os_mon_sysinfo can timeout.
    Causing crashes in mem_sup and disk_sup, every 2 seconds, leading
    to os_mon termination.

commit 1f768b7
Author:     Anders Nygren <anders.nygren@gmail.com>
AuthorDate: Thu Jun 30 16:42:09 2016 -0500
Commit:     Anders Nygren <anders.nygren@gmail.com>
CommitDate: Mon Aug 8 11:00:25 2016 -0500

    MB-19708: Prevent os_mon_sysinfo from returning stale data

    If get_data times out, any data received on the port would be
    added to the result of the next request. To avoid this the
    process exits, forcing a close and reopen of the port when
    the process is restarted.
erlang#1667

After internal discussion it was decided we need to enable wildcards
in SAN for Mad-Hatter hence the patch for erlang 20.

Erlang team added an option customize_hostname_check to enable SAN
wildcards in erlang 21 but we decided not to use their
solution because of the following reasons:
- appling solution from otp team would require extra work to be done
  in order to enable SAN wildcards for distribution (because they
  only provide an option, so distribution needs to start using this
  option);
- we will have to fix other then distribution clients that we use as
  well (mostly http clients I guess).

Related OTP commits:
erlang@f821c91
erlang@66ed8bc
In order to establish tls connection using raw IP address we should pass
connect address as a tuple instead of string. My understanding
is that it tries to verify it as a hostname otherwise.

Effectively same as commit by timofey-barmin, in couchbase-madhatter,
77ca6ab.
In order to enable logging tls keys on client connection in cb_dist, we
need access to the SSL socket, which is provided in inet_tls_dist by
calling cb_dist using the kernel application env.
This will allow us to more easily run tests in parallel if we don't have
to worry about hitting a 5 second timeout in any slightly long running
test.
Old value is 1 second which means net_kernel:stop() can return
after 1 second with distribution processes still running which
leads to inability to start net_kernel back immediately.

Hypothetically 60 seconds should be enough because it is more than
the sum of shutdown timeouts of all subprocesses.
I think we should not use infinity to make sure net_kernel:stop()
doesn't hang forever in case if we miss something here.
@stevewatanabe stevewatanabe merged commit 0fe80fa into couchbaselabs:dev Jan 25, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.