Mail app - FTS indexing fails on mailboxes with large attachments (HTTP timeout)

[paradoxbound]

With all due respect it has been almost 4 months since I reported this issue and almost 2 months since the last update. I don't care about new features in 10, I simply want the product I pay for to do what it says on the tin. I and my clients need to be able to search their mail.
I gave a fairly detailed bug report and I will happily supply more details if that is what is required.
At least give power users like myself the ability to edit our own configuration files.
Really not a happy bunny right now.

[girish]

@paradoxbound we are getting there with Cloudron 10. Unfortunately, there are no plans to make another release before that. Please hold on...

[girish]

@paradoxbound said:

Environment variable: DOVECOT_FTS_TIMEOUT (default: 60, recommended: 180-300 for large mailboxes)
Or alternatively: DOVECOT_FTS_BATCH_SIZE to reduce batch size from 200 to 50 which avoids calculating mailbox sizes.

I investigated this now finally. DOVECOT_FTS_TIMEOUT seems to be made up, there is no such thing. https://github.com/dovecot/core/blob/2.3.21/src/plugins/fts-solr/solr-connection.c#L105 hardcoded the timeout to 60 and this value is not configurable in dovecot.

The only thing we can do is to lower the batch size. I have made the change to cut this down to 50. @paradoxbound as a workaround since yo uare running this on QMEU, simply allocate more CPU power to the CLoudron VM to get over this for the moment. Or have you tried this already?

[girish]

@paradoxbound also to double check, do you see multiple errors with timeout ? Just wondering if dovecot moves on to other "batches" if one batch errors or not. According to various LLMs, it does move on, but I want to confirm this behavior in practice.

[paradoxbound]

I don't let LLMs on my production services but I do use them to help troubleshoot. Dovecot doesn't move on when a batch fails. I will rerun the troubleshooting and post some log output, along with actual config variables. I own those two incorrect variables. Claude has been spanked and I now have GLM 5.2 running as LLM-as-Judge on it's output.

[girish]

@paradoxbound No worries, I managed to reproduce the issue and have applied a fix. I have to reduce the batch size to 50. I have also provided a FTS status now on the dashboard to see the indexing status.

[necrevistonnezr]

On an updated Cloudron 10.0.1, I get „64“ mail indexing errors (lots of them):

Error indexing - Bulk: doveadm index -u user@domain.com, 
- Bulk failed with code 64

[girish]

@necrevistonnezr is there anything more in the mail container logs?

[girish]

Could it be you have a folder named "- Bulk" ?

[necrevistonnezr]

@girish said:

Could it be you have a folder named "- Bulk" ?
Yes, many of my folders have a - in their name so they sorted on top of folder lists.

From /home/yellowtent/platformdata/logs/mail/app.log⁠:

2026-08-26T06:50:10+02:00 [POST] /fts_config
2026-08-26T06:50:10+02:00 2026-08-26 04:50:10,632 INFO waiting for dovecot to stop
2026-08-26T06:50:11+02:00 2026-08-26 04:50:11,741 INFO stopped: dovecot (exit status 0)
2026-08-26T06:50:11+02:00 2026-08-26 04:50:11,745 INFO spawned: 'dovecot' with pid 7630
2026-08-26T06:50:11+02:00 Warning: fd limit (ulimit -n) is lower than required under max. load (1024 < 1500), because of default_client_limit
2026-08-26T06:50:12+02:00 2026-08-26 04:50:12,771 INFO success: dovecot entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
2026-08-26T06:50:11+02:00 2026-08-26 04:50:11,741 INFO stopped: dovecot (exit status 0)
2026-08-26T06:50:11+02:00 2026-08-26 04:50:11,745 INFO spawned: 'dovecot' with pid 7630
2026-08-26T06:50:11+02:00 Warning: fd limit (ulimit -n) is lower than required under max. load (1024 < 1500), because of default_client_limit
2026-08-26T06:50:12+02:00 2026-08-26 04:50:12,771 INFO success: dovecot entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
2026-08-26T06:50:21+02:00 2026-08-26 04:50:21,953 INFO spawned: 'solr' with pid 7877
2026-08-26T06:50:22+02:00 2026-08-26 04:50:22,011 WARN exited: solr (exit status 1; not expected)
2026-08-26T06:50:23+02:00 2026-08-26 04:50:23,014 INFO spawned: 'solr' with pid 7925
2026-08-26T06:50:24+02:00 2026-08-26 04:50:24,146 INFO success: solr entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
2026-08-26T06:50:24+02:00 2026-08-26 04:50:24,376 INFO spawned: 'tika' with pid 8014
2026-08-26T06:50:25+02:00 2026-08-26 04:50:25,931 INFO success: tika entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
[kdj@cloudron mail]$ tail app.log
2026-08-26T06:50:29+02:00 ==> Indexing mailbox - Bulk/Apartment for user: user@domain.com
2026-08-26T06:50:29+02:00 index: invalid option -- ' '
2026-08-26T06:50:29+02:00 ==> Failed to index mailbox - Bulk/Apartment for user: user@domain.com: doveadm index -u user@domain.com - Bulk/Apartment failed with code 64

[girish]

@necrevistonnezr thanks! found the issue.

[girish]

The core issue was a quoting problem. It was mistaking "-" as an argument instead of a folder name. Fixed in https://git.cloudron.io/platform/box/-/commit/04adf5de3470a944cf811a23613bceb9ea520b38