Anyone else seeing "Unix socket path is too long to generate an alternate address" on OpenLiteSpeed/SPanel?

A

Awack

New Member
#1
#1
Hi all I have a question regarding OLS

Running OpenLiteSpeed on a SPanel-managed VPS (Rocky Linux).
I am Getting this in the OLS error log continuously — every 20-90 seconds, on every domain on the box:

[WARN] Unix socket path is too long to generate an alternate address: /tmp/lshttpd/<domain>.sock.<absurdly long numeric suffix>

It correlates with suEXEC PHP worker respawns. Most of the time it's silent/harmless, but occasionally a real request lands during the gap and throws a 502/503/504 — had one confirmed full outage from it.

Questions for anyone running OLS + SPanel (or OLS elsewhere):
  • Have you seen this exact warning in your own error log?
  • Does it correlate with any intermittent 502/503/504s on your end?
  • Anyone found a cause or fix, or is this a known SPanel-specific behavior?

I am trying to find out if this is a known/common pattern or something specific to my setup. Appreciate any data points.

Awack
 
Cold-Egg

Cold-Egg

Administrator
#3
#3
Could you provide an example of the full socket path that contains the .sock.<absurdly long numeric suffix>?

For example, something like:
**/tmp/lshttpd/example.com.sock.123456789**

I'd like to check the actual length and format of the generated socket path.
 
A

Awack

New Member
#4
#4
Cold-Egg said:
Could you provide an example of the full socket path that contains the .sock.<absurdly long numeric suffix>?

For example, something like:
**/tmp/lshttpd/example.com.sock.123456789**

I'd like to check the actual length and format of the generated socket path.
Click to expand...


TY Cold-Egg,

__

Real example pulled directly from our error.log (OpenLiteSpeed + SPanel, Rocky Linux 10.2):

/tmp/lshttpd/depoteco.com.sock.819.136.394.641.169.060.300.139.574.532.429.945.425.583.793.086.056.245.078

That's 106 characters — right up against the 108-byte sun_path limit for AF_UNIX sockets on Linux (sizeof(sockaddr_un.sun_path)), which is why OLS can't append further and logs the "too long to generate an alternate address" WARN.

Format: fixed prefix /tmp/lshttpd/ (13 chars) + domain + .sock. + a dot-grouped numeric suffix that keeps growing — in our case it accumulated across repeated LSPHP worker recycles rather than being a single fixed-width value, until it hit the ceiling.

We ran this down with ScalaHosting support (citing OpenLiteSpeed's localworker.cpp/extworkerconfig.cpp): the alternate-socket suffix appends on every LSPHP recycle and only surfaces in the log once it actually fails — the base socket path itself is short and correctly SPanel-provisioned, it's not the constraint. Our mitigation: raised extMaxIdleTime (governs how long an idle LSPHP worker survives before being killed/recycled) from 30s to 3600s — cut the recycle frequency way down. WARN rate dropped from ~60/hr to 0 within about 90 minutes, holding clean since (2026-08-31).

TY for look at this.. its drving me and my Sysops crazy :)
 
You must log in or register to reply here.
Top