Health and checks¶

lanstash checks itself: the Health page and lanstash doctor say whether the store loaded cleanly and whether open files and disk space are fine, every stored byte carries a checksum that a full check can read back, and eviction keeps the disk from filling up. This page covers all of these.

The Health page¶

Health in the sidebar lists the checks lanstash makes of itself, each ok, warning or failure, with a summary on top (all good, something to look at, something is wrong). The Overview's status line shows the same result.

lanstash doctor¶

lanstash doctor checks the installation from outside, read-only:

lanstash doctor

It looks at the version, the service's state, the configuration, admin_listen, each listen port, the admin API, open files against the service's limit, each store root (writable, and its free space against keep_free), the store's contents (with the background check's problems), the upstream DNS, the cache-domains lists and the log directories. It prints one line per check ([ok ], [WARN] or [FAIL]), then the counts, and exits with 1 when any check fails.

Checksums¶

Every stored byte has a checksum: each small object in a segment file, and each range of a large object file. Downloads are checked before they are kept where the platform allows it: Steam chunks against the CRC Steam sends, Epic chunks and Battle.net files against the hashes in their manifests.

lanstash store verify¶

lanstash store verify reads every stored byte and checks it against its checksum. It only reads, so it is safe while lanstash serves.

lanstash store verify                       # as fast as the disks go, 4 threads
lanstash store verify --rate 100            # at most 100 MB a second
lanstash store verify --threads 8           # more threads
lanstash store verify --accept              # do not ask first (systemd, scripts)

Before it starts, it says how much it will read and asks you to go on. Without a terminal to ask (under systemd, or in a script) it starts only with --accept. It reports damaged records, unreadable files and journal problems, and exits with 1 when it finds any. It repairs nothing: a damaged object is fetched again from the internet when a client asks for it.

The background check¶

With Check stored bytes (verify_every) set in Settings, lanstash runs the same check slowly in the background while it serves: weekly, monthly, every n days (<n>d), or off (the default).

• Check speed (verify_rate) caps how much it reads a second (20M, 50M, 100M; 50M by default), so clients keep the disks.
• It resumes where it was after a restart.
• Whatever looks wrong is checked again against a fresh listing before it is reported, so a file deleted meanwhile is not reported as damage.
• The Overview's Storage page shows the Store check: when it last finished, how far a running round got, when the next one is due, and any Damage found. lanstash's log names each damaged record or file.

Eviction¶

When the store runs short of room, lanstash deletes what was used least recently. Two limits, both optional, set in Settings (the Storage group) or the configuration; the stricter one wins:

• Eviction works on whole units: a segment file of small objects, or one large object file. The segment being written is never evicted, and anything used in the last 10 minutes is protected.
• Nothing is evicted for its age alone: game downloads are named by their content and never go stale, so lanstash keeps them as long as there is room.
• The Storage page shows what was Evicted, with Segments evicted, Files evicted and any Eviction failures.

Settings changed on the web UI apply when lanstash restarts; Settings offers Restart for that.