Server Version#: 1.43.2.10687 (proxmox helper scripts version - ubuntu)
Player Version#:10.30.8.4222 android / google tv
<If providing server logs please do NOT turn on verbose logging, only debug logging should be enabled>
Hey everyone,
My Plex server intermittently experiences severe database lockups (about once a week), resulting in client timeouts and Broken pipe errors.
My Environment:
-
Server: Plex Media Server in a Proxmox LXC (Proxmox Helper Scripts).
-
DB Storage: Local SSD (Main DB is 259 MB, Blobs is 384 MB). DB integrity is ok.
-
Media Storage: Unraid server array, mounted via SMB on the Proxmox host and passed through to the LXC.
The Issue: Out of nowhere a local client will try to start playback or update a subtitle tracking state, and SQLite completely deadlocks. During the lockup, background timeline pings from a remote user who is currently sitting on a paused transcode session pile up behind the write-lock, compounding the traffic jam.
Here is the sequence from the logs during the crash:
// Local client requests subtitle stream change, opening a transaction
Jun 05, 2026 20:34:06.978 DEBUG - Request: [192.168.1.214:44598] PUT /library/parts/100548?subtitleStreamID=446785&allParts=1…
// SQLite immediately struggles to close the transaction
Jun 05, 2026 20:34:08.398 WARN - [Req#1eb4f0] Held transaction for too long: 0.240000 seconds
// Remote user (paused on a WAN connection) attempts a routine timeline check-in, getting stuck behind the lock
Jun 05, 2026 20:34:18.204 DEBUG - Request: [192.168.1.122:50106 (WAN)] GET /:/timeline?.. state=paused
// The initial subtitle query finally clears after a massive 26-second delay
Jun 05, 2026 20:34:21.149 WARN - [Req#1eb4f0] SLOW QUERY: It took 26420.000000 ms to retrieve 1 items.
// The database queue cascades, choking subsequent metadata requests
Jun 05, 2026 20:34:26.366 WARN - [Req#1eb543] SLOW QUERY: It took 25350.000000 ms to retrieve 1 items.
// The local client times out and drops the connection entirely
Jun 05, 2026 20:34:37.958 DEBUG - Failed to stream media, client probably disconnected after 1540096 bytes: 32 - Broken pipe
// The remote paused user’s request finally resolves after being trapped for 22 seconds
Jun 05, 2026 20:34:48.636 WARN - [Req#1eb527] SLOW QUERY: It took 22520.000000 ms to retrieve 2 items.
Given that the DB is on local NVMe, could a minor latency hiccup over the host-level Unraid SMB media mount cause a simple PUT /library/parts subtitle update to hang like this and deadlock the database? What should I check next?
