CEPH-OSD(8)

📦
version 20.2.4 · 20 versions
, in ubuntu stonking

💖 CEPH-OSD(8) 💖

okay bestie, welcome to the ceph-osd page from the Ceph manual — yes that Ceph 💅. this is the whole thing, every flag, every default, no skips, no shortcuts. grab a snack 🍓

✨ NAME ✨

ceph-osd — the ceph object storage daemon 🫶

✨ SYNOPSIS ✨

ceph-osd -i osdnum [ --osd-data datapath ] [ --osd-journal
journal ] [ --mkfs ] [ --mkjournal ] [--flush-journal] [--check-allows-journal] [--check-wants-journal] [--check-needs-journal] [ --mkkey ] [ --osdspec-affinity ]

like, that's the shape of it. -i osdnum is the main character, everything else is optional and tagged on 💅

✨ DESCRIPTION ✨

so ceph-osd is the object storage daemon for the Ceph distributed file system 🦋. it manages data on local storage with redundancy, and then it serves that data out over the network. it's giving "I hold your bytes AND I share them" energy, honestly iconic.

for Filestore-backed clusters, the argument of the --osd-data datapath option (and yeah, datapath is literally datapath in this example) should be a directory on an XFS file system where the object data lives. that's the spot. that's home 🏠

the journal? optional. periodt. but here's the thing — the journal only improves performance when it lives on a different disk than the disk specified by datapath. if it's on the same disk, don't bother, girl. and the storage medium the journal sits on should be a low-latency medium — ideally an SSD device. no cap, spinning rust is not it for a journal 🔥

✨ OPTIONS ✨

okay this is the part everyone actually came for. every single one, no gatekeeping 💜

  • -f, --foreground — foreground mode: do not daemonize after startup (so it just runs in the foreground, right there, in front of you 👀). it also does not generate a pid file. super useful when you're running it via ceph-run(8).
  • -d — debug mode: it's like -f, but it also sends all log output to stderr. chaos mode, but make it informative 💀
  • --setuser userorgid — set the uid after starting. if you hand it a username, the user record gets looked up to grab a uid and a gid, and the gid gets set too — unless --setgroup is also specified. so like, they talk to each other, besties 🤝
  • --setgroup grouporgid — set the gid after starting. if you hand it a group name, the group record gets looked up to grab a gid. simple as 🎀
  • --osd-data osddata — use the object store at osddata.
  • --osd-journal journal — journal the updates to journal.
  • --check-wants-journal — check whether a journal is desired 🙈
  • --check-allows-journal — check whether a journal is allowed ✅
  • --check-needs-journal — check whether a journal is required 😤
  • --mkfs — create an empty object repository. this also initializes the journal (if one is defined). the big "start from nothing" button ✨
  • --mkkey — generate a new secret key 🔑. this is normally used in combination with --mkfs, because honestly it's way more convenient than generating a key by hand with ceph-authtool(8).
  • --mkjournal — create a new journal file so it matches an existing object repository. majorly useful when the journal device or file got wiped out by a disk or file system failure 😭 (we've all been there)
  • --flush-journal — flush the journal to permanent store. this runs in the foreground, so you actually know when it's finished 🫡. useful if you want to resize the journal or otherwise need to destroy it: this guarantees you won't lose data 💖
  • --get-cluster-fsid — print the cluster fsid (uuid) and exit.
  • --get-osd-fsid — print the OSD's fsid and exit. the OSD's uuid is generated at --mkfs time, so it's unique to a particular instantiation of this OSD 👑
  • --get-journal-fsid — print the journal's uuid. the journal fsid gets set to match the OSD fsid at --mkfs time. they're matching, it's cute 🌸
  • -c ceph.conf, --conf=ceph.conf — use the ceph.conf configuration file instead of the default /etc/ceph/ceph.conf for runtime configuration options.
  • -m monaddress[:port] — connect to the specified monitor (instead of looking through ceph.conf).
  • --osdspec-affinity — set an affinity to a certain OSDSpec. this option can only be used in conjunction with --mkfs. no --mkfs, no deal 💅

✨ AVAILABILITY ✨

ceph-osd is part of Ceph, a massively scalable, open-source, distributed storage system 🌈. hit up the Ceph documentation at https://docs.ceph.com for more info, iykyk.

✨ SEE ALSO ✨

ceph(8), ceph-mds(8), ceph-mon(8), ceph-authtool(8) — the whole friend group fr fr 🫶

2010-2014, Inktank Storage, Inc. and contributors. Licensed under Creative Commons Attribution Share Alike 3.0 (CC-BY-SA-3.0) 🎀

okay and that's the whole page, bestie 🦋✨ this one's from the dev version of the Ceph manual, dated Oct 06, 2026, and yeah — it's still filed under that CEPH-OSD(8) name 💜 go run your OSD, you've got this 👑

📦 where this page comes from