Plugin store mongo_replset: got unrecoverable error in primary and no secondary error_class=Encoding::UndefinedConversionError error="\"\\xC3\" from ASCII-8BIT to UTF-8"

[quoccuongmmt]

My config td-agent

  <store>
    @type mongo_replset
    <inject>
      time_key updated_at
    </inject>

    nodes 1.2.3.4:27017,1.2.3.5:27017,1.2.3.6:27017
    replica_set abc
    database abc
    collection logs_%m_%Y
    user ...
    password ....
    date_keys created_at,request_time

    <buffer tag, time>
      @type file
      path /var/log/td-agent/buffer/mongodb
      timekey 60
      timekey_wait 60
    </buffer>

  </store>

Some log code have special-character word like {"character": "༂Sυɠαɾ ɗαɗү࿐"}. This log can't push to mongo and I was checked file log td-agent.log:

2020-09-18 09:58:02 +0700 [warn]: #0 got unrecoverable error in primary and no secondary error_class=Encoding::UndefinedConversionError error="\"\\xE1\" from ASCII-8BIT to UTF-8"
2020-09-18 09:58:02 +0700 [warn]: #0 suppressed same stacktrace
2020-09-18 09:58:02 +0700 [warn]: #0 bad chunk is moved to /tmp/fluent/backup/worker0/object_3fba0e62a1a8/5af8da4b100f9f43f77fac346c1c4274.log

I'm using OS: Centos 7 td-agent 3 Can everyone help me?

[repeatedly]

Check our FAQ. There are several approaches for it:

https://docs.fluentd.org/quickstart/faq#i-got-encoding-error-inside-the-plugin-how-to-fix-it

[djerveren]

Check our FAQ. There are several approaches for it:

https://docs.fluentd.org/quickstart/faq#i-got-encoding-error-inside-the-plugin-how-to-fix-it

Hello. Sorry for the necro, but I figured it'd be better to posting in an already existing issue than creating a new one on the same topic.

We are facing a similar error as @quoccuongmmt: #0 got unrecoverable error in primary and no secondary error_class=Encoding::UndefinedConversionError error="\"\\xC3\" from ASCII-8BIT to UTF-8"

I had a look at your link there @repeatedly and tried the record_modifier solution, but sadly it didn't help (or quite possibly, I didn't use it correctly). We still get the same error after a while, and we have a backup directory filling up with bad chunks in files.

These files contain lines like this:

<92>Îf8g«<84>Ä^KdefaultData<86>Ä^Guser_idÎ^@^AU-Ä^Nsecurity_level<91>Ä^CpwdÄ^Mactive_idÀÄ^FactionÄ^QLogin: SuccessfulÄ
activeYearÍ^GèÄ^KactiveMonth^EÄ^Eextra<82>Ä^HusernameÄ^LMarOskuerrÄ^Kauth_methodÄ^HpasswordÄ^ElevelÄ^DINFO¤dist¤app<92>Îf8gî<84>Ä^KdefaultData<86>Ä^Guser_idÎ^@^Cµ|Ä^Nsecurity_level<91>Ä^CpwdÄ^Mactive_idÀÄ^FactionÄ^QLogin: SuccessfulÄ

This is our config:

<system>
  log_level info
</system>

<source>
  @type monitor_agent
  bind 0.0.0.0
  port 24220
</source>

<source>
  @type  forward
  @id    app
  @label @app
  port  24224
</source>

<label @app>
    <match *.action.*>
      @type mongo
      host mongo_host
      port 27017
      database mongo_db
      collection log_action
      user mongo_user
      password mongo_pwd

      <buffer>
        flush_interval 10s
      </buffer>
      <inject>
        time_key time
      </inject>
    </match>

    <match *.user.*>
      @type mongo
      host mongo_host
      port 27017
      database mongo_db
      collection log_user
      user mongo_user
      password mongo_pwd

      <buffer>
        flush_interval 10s
      </buffer>
      <inject>
        time_key time
      </inject>
    </match>
</label>

I tried two things with record_modifier where I first put the following before the <label @app> section:

...
<source>
  @type  forward
  @id    app
  @label @app
  port  24224
</source>

<filter **>
    @type record_modifier
    char_encoding utf-8
</filter>

<label @app>
    <match *.action.*>
      @type mongo
      host mongo_host
...

When that didn't work, I tried the following instead:

...
<source>
  @type  forward
  @id    app
  @label @app
  port  24224
</source>

<filter **>
    @type record_modifier
    char_encoding ascii-8bit:utf-8
</filter>

<label @app>
    <match *.action.*>
      @type mongo
      host mongo_host
...

But to no avail. Maybe I'm not understanding the instructions in the in the record_modifier README or simply misunderstand how fluentd works (I am not all that familiar with the software). We would greatly appreciate any help/pointers for solving this issue.

OS: Ubuntu 22.04 td-agent 4.4.2 fluentd 1.15.3 gem fluent-plugin-mongo (1.6.0) gem mongo (2.18.2)

[christiansipola]

I wonder if there should be a .force_encoding "utf-8" added somewhere in the code, similar to https://github.com/SpringMT/fluent-plugin-ses/pull/1/