Add a patch for RubyInstaller to avoid crash on start up #620

ashie · 2024-02-16T05:39:42Z

When a non-ASCII key exists under the registry key SOFTWARE/Microsoft/Windows/CurrentVersion/Uninstall/, Fluentd fails to start workers due to Encoding::UndefinedConversionError. This patch avoid this issue.

Fix #616

ashie · 2024-02-16T08:22:55Z

Hmm, this patch breaks searching MSYS2 for build 🤔

fluent-package/msi/build.bat

ashie · 2024-02-20T09:39:47Z

I've confirmed that this patch fixes the issue.

When a non-ASCII key exists under the registry key `SOFTWARE/Microsoft/Windows/CurrentVersion/Uninstall/`, Fluentd fails to start workers due to `Encoding::UndefinedConversionError`. This patch avoid this issue. Fix #616 Signed-off-by: Takuro Ashie <[email protected]>

ashie · 2024-02-22T05:17:16Z

We observed that supervisor process of fluentd is finished unexpectedly after about 1 hour passed while repeating recovery. While this situation, opened handles are continually increased, over than 8400 at last.

We got following backtrace on finishing supervisor process.

2024-02-21 15:34:23 +0900 [debug]: fluent/log.rb:341:debug: Got Win32 event "fluentd_7928_STOP_EVENT_THREAD"
Unexpected error undefined method `pid' for nil:NilClass
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/fluentd-1.16.3/lib/fluent/supervisor.rb:417:in `after_start'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_spawn_server.rb:77:in `ensure in start_worker'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_spawn_server.rb:77:in `start_worker'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:175:in `delayed_start_worker'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:159:in `restart_worker'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:125:in `block in keepalive_workers'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:102:in `each'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:102:in `each_with_index'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:102:in `keepalive_workers'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:58:in `run'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_spawn_server.rb:50:in `run'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/server.rb:128:in `main'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/daemon.rb:119:in `main'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/daemon.rb:68:in `run'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/fluentd-1.16.3/lib/fluent/supervisor.rb:796:in `supervise'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/fluentd-1.16.3/lib/fluent/supervisor.rb:582:in `run_supervisor'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/fluentd-1.16.3/lib/fluent/command/fluentd.rb:352:in `<top (required)>'
  <internal:C:/opt/fluent/lib/ruby/3.2.0/rubygems/core_ext/kernel_require.rb>:85:in `require'
  <internal:C:/opt/fluent/lib/ruby/3.2.0/rubygems/core_ext/kernel_require.rb>:85:in `require'
  C:/opt/fluent/lib/ruby/gems/3.2.0/gems/fluentd-1.16.3/bin/fluentd:15:in `<top (required)>'
  C:/opt/fluent/bin/fluentd:32:in `load'
  C:/opt/fluent/bin/fluentd:32:in `<main>'

In the above log, the root cause is squashed by ensure. So I fetched additional backtrace of the exception:

#<Errno::EMFILE: Too many open files - dup>
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/process_manager.rb:190:in `spawn'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/fluentd-1.16.3/lib/fluent/supervisor.rb:413:in `spawn'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_spawn_server.rb:75:in `start_worker'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:175:in `delayed_start_worker'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:159:in `restart_worker'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:125:in `block in keepalive_workers'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:102:in `each'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:102:in `each_with_index'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:102:in `keepalive_workers'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_worker_server.rb:58:in `run'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/multi_spawn_server.rb:50:in `run'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/server.rb:128:in `main'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/daemon.rb:119:in `main'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/serverengine-2.3.2/lib/serverengine/daemon.rb:68:in `run'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/fluentd-1.16.3/lib/fluent/supervisor.rb:796:in `supervise'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/fluentd-1.16.3/lib/fluent/supervisor.rb:582:in `run_supervisor'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/fluentd-1.16.3/lib/fluent/command/fluentd.rb:352:in `<top (required)>'
<internal:C:/opt/fluent/lib/ruby/3.2.0/rubygems/core_ext/kernel_require.rb>:85:in `require'
<internal:C:/opt/fluent/lib/ruby/3.2.0/rubygems/core_ext/kernel_require.rb>:85:in `require'
C:/opt/fluent/lib/ruby/gems/3.2.0/gems/fluentd-1.16.3/bin/fluentd:15:in `<top (required)>'
C:/opt/fluent/bin/fluentd:32:in `load'
C:/opt/fluent/bin/fluentd:32:in `<main>'

ashie · 2024-02-22T05:30:20Z

We observed that supervisor process of fluentd is finished unexpectedly after about 1 hour passed while repeating recovery.

I filed a new issue for tracking this problem at ServerEngine's repository:
treasure-data/serverengine#145

kenhys · 2024-02-22T05:36:28Z

Before:

crashed because of problematic registry.

After:

uninstalled msys64 beforehand.
No crash on startup.

ashie · 2024-02-22T05:37:32Z

バッファローらくらくアップデート！お前か！

kenhys

LGTM

When a non-ASCII key exists under the registry key `SOFTWARE/Microsoft/Windows/CurrentVersion/Uninstall/`, Fluentd fails to start workers due to `Encoding::UndefinedConversionError`. This patch avoid this issue. Backported from v5.0.3: #620 Signed-off-by: Takuro Ashie <[email protected]>

When a non-ASCII key exists under the registry key `SOFTWARE/Microsoft/Windows/CurrentVersion/Uninstall/`, Fluentd fails to start workers due to `Encoding::UndefinedConversionError`. This patch avoid this issue. Backported from v5.0.3: #620 --------- Signed-off-by: Takuro Ashie <[email protected]>

daipom · 2024-04-24T04:17:29Z

note: Avoid crash even if a registry key incldues inconvertible characters oneclick/rubyinstaller2#375 has been merged.

ashie force-pushed the avoid-crash-on-non-ascii-registry branch from 0bb59ad to 3081471 Compare February 16, 2024 05:41

daipom self-requested a review February 16, 2024 05:41

ashie force-pushed the avoid-crash-on-non-ascii-registry branch 3 times, most recently from b7d7c3d to 592fc53 Compare February 16, 2024 07:31

ashie force-pushed the avoid-crash-on-non-ascii-registry branch 2 times, most recently from ff55475 to 7c937a3 Compare February 16, 2024 09:43

ashie commented Feb 16, 2024

View reviewed changes

fluent-package/msi/build.bat Outdated Show resolved Hide resolved

ashie force-pushed the avoid-crash-on-non-ascii-registry branch from 7c937a3 to c4b4056 Compare February 20, 2024 03:37

ashie marked this pull request as ready for review February 20, 2024 09:39

ashie added this to the 5.0.3 milestone Feb 20, 2024

ashie requested a review from kenhys February 21, 2024 06:09

ashie force-pushed the avoid-crash-on-non-ascii-registry branch from c4b4056 to f651dac Compare February 22, 2024 01:10

ashie mentioned this pull request Feb 22, 2024

Suspected resource leak in worker recovery feature treasure-data/serverengine#145

Open

kenhys approved these changes Feb 22, 2024

View reviewed changes

kenhys merged commit 8a17a8c into master Feb 22, 2024
22 checks passed

kenhys deleted the avoid-crash-on-non-ascii-registry branch February 22, 2024 07:04

ashie mentioned this pull request Feb 29, 2024

td-agent-v4: Backport #620 #625

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add a patch for RubyInstaller to avoid crash on start up #620

Add a patch for RubyInstaller to avoid crash on start up #620

ashie commented Feb 16, 2024

ashie commented Feb 16, 2024

ashie commented Feb 20, 2024

ashie commented Feb 22, 2024 •

edited

Loading

ashie commented Feb 22, 2024

kenhys commented Feb 22, 2024 •

edited

Loading

ashie commented Feb 22, 2024

kenhys left a comment

daipom commented Apr 24, 2024

Add a patch for RubyInstaller to avoid crash on start up #620

Add a patch for RubyInstaller to avoid crash on start up #620

Conversation

ashie commented Feb 16, 2024

ashie commented Feb 16, 2024

ashie commented Feb 20, 2024

ashie commented Feb 22, 2024 • edited Loading

ashie commented Feb 22, 2024

kenhys commented Feb 22, 2024 • edited Loading

ashie commented Feb 22, 2024

kenhys left a comment

Choose a reason for hiding this comment

daipom commented Apr 24, 2024

ashie commented Feb 22, 2024 •

edited

Loading

kenhys commented Feb 22, 2024 •

edited

Loading