Hello,
I’m running the trial edition of LiteSpeed Web Server Enterprise on a DirectAdmin / AlmaLinux 8 server and have run into a repeatable memory issue that takes every virtual host offline.
Environment
What happens
I’m running the trial edition of LiteSpeed Web Server Enterprise on a DirectAdmin / AlmaLinux 8 server and have run into a repeatable memory issue that takes every virtual host offline.
Environment
- OS : AlmaLinux 8.8
- Control panel : DirectAdmin 1.66.4, CustomBuild 2.0
- LSWS version : LiteSpeed Web Server Enterprise (trial)
- PHP handlers : Multi php 7.4, 8.2, 5.6 built via CustomBuild
- Server RAM : 64 GB
What happens
- After several hours of normal operation, memory_limit inside lsphp suddenly drops from the configured 8096M to 512M.
- PHP quickly exhausts that limit and throws
PHP Fatal error: Out of memory (allocated 627539968) (tried to allocate 8192 bytes) - After some time who i tried to solve the memory issue (many website refresh), all sites then return 503 Service Unavailable.
2025-06-24 09:28:29.479577 [INFO] [526562] [T0] [...#APVH_www.tts.gr:443]
connection to [uds://usr/local/lsws/extapp-sock/APVH_tts_Suphp74.sock] …
error: No such file or directory!
2025-06-24 09:28:29.479624 [NOTICE] … Max retries has been reached, 503!
2025-06-24 09:28:29.479639 [NOTICE] … 503 Service Unavailable
2025-06-24 09:28:30.954623 [INFO] … error: Connection refused!
What I have verified/tried
- memory_limit = 8096M in /usr/local/php74/lib/php.ini and confirmed via phpinfo() before the issue appears.
- No .htaccess, pool, or per-virtual-host overrides exist.
- Plenty of free RAM; no OOM-killer activity in dmesg or journal.
- Cleared LSWS opcode & public caches, restarted the LiteSpeed service, and performed a full server reboot — the 503 error persisted every time.
Questions- Are you aware of any bug or trial-edition limitation that could force a 512 MB memory_limit?
- Does LSWS or the DirectAdmin integration have any hidden per-user or server-wide cap that could explain this?
- Which additional logs or debug flags would you like me to enable to help pinpoint the source of the override?
Any guidance or a workaround would be greatly appreciated, as this is a production-bound migration test.
Thank you in advance for your help.