The OPcache buffer is nearly full no matter the actual setting

[tkopicbarrage]

I think i may have found the bug, I am running: nextcloud:31.0.6-fpm

It's running in my kubernetes cluster on aarch64 machines, deployed via the official helm chart.

it's been running inside this same cluster for years now, ever since nextcloud 17.
Everything is setup to auto upgrade as soon as the newer image is available and I have been getting this fake opcache full warning ever since 30.0.0-fpm came out.
I tried to tweak the configuration multiple times but i had no luck so far so i decided to open up this issue.

My current php pool looks like this:

[www]                                                                                                                                                                                                                                                                                    
user = www-data                                                                                                                                                                                                                                                                          
group = 82                                                                                                                                                                                                                                                                               
listen = 127.0.0.1:9000                                                                                                                                                                                                                                                                  
pm = dynamic                                                                                                                                                                                                                                                                             
pm.max_children = 64                                                                                                                                                                                                                                                                     
pm.start_servers = 16                                                                                                                                                                                                                                                                     
pm.min_spare_servers = 1                                                                                                                                                                                                                                                                 
pm.max_spare_servers = 16                                                                                                                                                                                                                                                                 
env[TMP] = /var/www/tmp                                                                                                                                                                                                                                                                  
env[TMPDIR] = /var/www/tmp                                                                                                                                                                                                                                                               
env[TEMP] = /var/www/tmp                                                                                                                                                                                                                                                                 
php_admin_value[upload_tmp_dir] = /var/www/tmp                                                                                                                                                                                                                                           
php_admin_value[max_execution_time] = 3600                                                                                                                                                                                                                                               
php_admin_value[max_input_time] = 3600                                                                                                                                                                                                                                                   
php_admin_value[upload_max_filesize] = 4G                                                                                                                                                                                                                                                
php_admin_value[post_max_size] = 4G                                                                                                                                                                                                                                                      
php_admin_value[memory_limit] = 4096M                                                                                                                                                                                                                                                    
php_admin_value[opcache.save_comments] = 1                                                                                                                                                                                                                                               
php_admin_value[opcache.interned_strings_buffer] = 1024                                                                                                                                                                                                                                  
php_admin_value[opcache.memory_consumption] = 2048                                                                                                                                                                                                                                       
php_admin_value[opcache.max_accelerated_files] = 30000                                                                                                                                                                                                                                   
php_admin_value[opcache.validate_timestamps] = 0                                                                                                                                                                                                                                         
php_admin_value[opcache.revalidate_freq] = 120                                                                                                                                                                                                                                           
php_admin_value[opcache.jit] = 1255                                                                                                                                                                                                                                                      
php_admin_value[opcache.jit_buffer_size] = 256M  

I've been getting these warnings no matter what number i set opcache.memory_consumption to:
For example:

The warning is visible as soon as the container starts up, the default configs recommended by the official documentation and the ones provided in the helm chart values also return the same warning obviously as the default value is 256M.

Edit:
the Pod itself has 3 containers, nextcloud, nginx and nextcloud-cron and it uses only about 500mb of memory on average, peaks only at about 2GB when intensively used.

{
    "system": {
        "memcache.local": "\\OC\\Memcache\\APCu",
        "apps_paths": [
            {
                "path": "\/var\/www\/html\/apps",
                "url": "\/apps",
                "writable": false
            },
            {
                "path": "\/var\/www\/html\/custom_apps",
                "url": "\/custom_apps",
                "writable": true
            }
        ],
        "overwriteprotocol": "https",
        "overwrite.cli.url": "https:\/\/cloud.kopic.hr",
        "maintenance_window_start": "23",
        "jpeg_quality": "75",
        "enable_previews": true,
        "enabledPreviewProviders": [
            "OC\\Preview\\BMP",
            "OC\\Preview\\GIF",
            "OC\\Preview\\JPEG",
            "OC\\Preview\\SVG",
            "OC\\Preview\\MarkDown",
            "OC\\Preview\\MP3",
            "OC\\Preview\\Movie",
            "OC\\Preview\\PNG",
            "OC\\Preview\\TXT",
            "OC\\Preview\\XBitmap",
            "OC\\Preview\\HEIC"
        ],
        "preview_max_x": "1280",
        "preview_max_y": "1280",
        "preview_max_scale_factor": 10,
        "preview_max_memory": 1024,
        "preview_concurrency_all": 16,
        "preview_concurrency_new": 4,
        "default_phone_region": "HR",
        "trusted_proxies": "***REMOVED SENSITIVE VALUE***",
        "memcache.distributed": "\\OC\\Memcache\\Redis",
        "memcache.locking": "\\OC\\Memcache\\Redis",
        "redis": {
            "host": "***REMOVED SENSITIVE VALUE***",
            "port": "6379"
        },
        "objectstore": {
            "class": "OC\\Files\\ObjectStore\\S3",
            "arguments": {
                "bucket": "nextcloud",
                "autocreate": false,
                "key": "***REMOVED SENSITIVE VALUE***",
                "secret": "***REMOVED SENSITIVE VALUE***",
                "hostname": "node5-minio.minio",
                "port": 9000,
                "region": "us-east-1",
                "use_path_style": true,
                "legacy_auth": false,
                "use_ssl": false
            }
        },
        "passwordsalt": "***REMOVED SENSITIVE VALUE***",
        "secret": "***REMOVED SENSITIVE VALUE***",
        "trusted_domains": [
            "localhost",
            "cloud.kopic.hr"
        ],
        "datadirectory": "***REMOVED SENSITIVE VALUE***",
        "dbtype": "pgsql",
        "version": "31.0.6.2",
        "dbname": "***REMOVED SENSITIVE VALUE***",
        "dbhost": "***REMOVED SENSITIVE VALUE***",
        "dbport": "",
        "dbtableprefix": "oc_",
        "dbuser": "***REMOVED SENSITIVE VALUE***",
        "dbpassword": "***REMOVED SENSITIVE VALUE***",
        "installed": true,
        "instanceid": "***REMOVED SENSITIVE VALUE***",
        "mail_smtpmode": "smtp",
        "mail_smtphost": "***REMOVED SENSITIVE VALUE***",
        "mail_smtpport": "465",
        "mail_smtpauth": true,
        "mail_smtpauthtype": "PLAIN",
        "mail_smtpname": "***REMOVED SENSITIVE VALUE***",
        "mail_smtppassword": "***REMOVED SENSITIVE VALUE***",
        "mail_from_address": "***REMOVED SENSITIVE VALUE***",
        "mail_domain": "***REMOVED SENSITIVE VALUE***",
        "mail_sendmailmode": "smtp",
        "dbpersistent": true,
        "maintenance": false,
        "mail_smtpsecure": "ssl",
        "loglevel": 3,
        "app_install_overwrite": [],
        "theme": "",
        "data-fingerprint": "27d6e54f3cc1968bdbd7009251646ccf"
    }
}

[joshtrichards]

Possibly related: #53429 (but only in v32 currently)

[joshtrichards]

Can you post the output of the opcache section (stats and settings) from phpinfo accessed via Admin settings->System->PHP->Show phpinfo

You may need to enable it first: https://github.com/nextcloud/serverinfo/?tab=readme-ov-file#show-phpinfo--nextcloud-28

[tkopicbarrage]

[MichaIng]

Can you check via https://github.com/amnuts/opcache-gui/blob/master/index.php which files do actually consume such insane amount of OPcache space? I tend to believe it is more a bug in the particular PHP build/implementation, either in the way the OPcache is used, or only regarding the status output.

Would be useful to add JIT usage to https://github.com/nextcloud/serverinfo. I believe in this case, with tracing JIT, only ~1 MiB of the 256 MiB are used, while everything else is either allocated from base OPcache memory or not at all, I mean the overall data size of Nextcloud is below 2 GiB. I never saw an instance with tracing JIT which ever showed any different value than ~1 MiB usage, regardless whether Nextcloud only, or a buch of other websites, or Wordpress + Matomo or whatever, always the same (after all the "hot" thresholds have been reached). Only when disabling tracing, and just compiling all scripts on first usage (e.g. opcache.jit=1205), declared JIT buffer is actually used.

One thing: the internet strings buffer is set to 1 GiB, which is insane, and takes away that 1 GiB from overall OPcache memory right from the start. The remaining 1 GiB should be still a magnitude of order above what Nextcloud takes. Nextcloud only shows interned strings buffer recommendations (if seen nearly full) if it is smaller than 25% of overall OPcache size. And currently, which is also weird, usage stats show 17 MiB of 32 MiB (!) interned strings buffer used, while the setting below shows 1024 MiB assigned. I checked my instances, and used + free interned strings buffer always match the 16 MiB size I gave it, down to the last byte.

So this is a K10s implementation of Nextcloud. I see it uses Nginx, but not how PHP is implemented. Is it a common dedicated PHP-FPM server? I know php_admin_value directive from Apache with mod_php only, but this container has a different way to add/change PHP settings, which seems to create common php.ini drop-ins: https://artifacthub.io/packages/helm/nextcloud/nextcloud#adjusting-php-ini-values
Maybe an idea to switch to that method and see whether it makes a difference. These contradicting interned strings buffer sizes and insane OPcache usage at least indicated that currently something is odd.

[tkopicbarrage]

I don't think it actually uses that much memory and this seems to be the false status output.

My kubernetes nodes have enough memory to handle this and i don't mind higher memory usage but the whole pod is only using about 800MB of ram, here's the screenshot of all the containers inside the pod:

oh, and about the strings buffer, I've played around with it and i just forgot to change it back, i usually keep it at 64

i mean this cannot be real, look at this if i set:
php_admin_value[opcache.memory_consumption] = 65535

phpinfo-2025-07-07_07.45.52.mp4

You can see the opcache reports to be full as soon as the container boots up,
the PHP container was only peaking out at about 220mb of memory

[MichaIng]

I think pushing OPcache settings values via php_admin_value is not working properly. I remember we tried this in attempt to apply a certain value for a certain sub directory only, but it does not work like that: it is applied for the whole OPcache instance used. But it also caused weird inconsistencies with Apache mod_php, showing weird values, varying on status page reload by times, since every Apache process included an on OPcache instance, some with the pushed values applied, some not. I am not sure how pushing such via PHP-FPM config works, but I assume it has similar issues as the value is not set once for the whole PHP-FPM master process, but on pool spawn, maybe even on access, during operation. A result might be that status shows the size that was set via php_admin_value, but with a usage percentage based on the size set via php.ini originally.

As said, do not use php_admin_value, but apply PHP settings properly via php.ini, as explained here: https://artifacthub.io/packages/helm/nextcloud/nextcloud#adjusting-php-ini-values