object-server.conf(5)

📦
version 2.38.2 · 3 versions
, in debian trixie, nixos 26.05, nixpkgs unstable, ubuntu questing, ubuntu resolute, ubuntu stonking

✨ NAME ✨

okay bestie, say hi to object-server.conf — the configuration file for the OpenStack Swift object server. This manual page is object-server.conf(5), and it's giving “I run your object storage vibes” 💅

🎀 SYNOPSIS 🎀

object-server.conf

That's it. A whole synopsis that's just the filename. iconic, honestly.

💖 DESCRIPTION 💖

This is the config file used by the object server and other object background services, like: replicator, reconstructor, updater, auditor, and expirer. They all pull up to this file.

The file follows python-pastedeploy syntax. It's split into sections wrapped in square brackets []. Each section has some number of key/value parameters, and we're gonna spill all the tea on those later.

Any line starting with a # symbol is ignored, so go off with comments 💬

Wanna know more about python-pastedeploy config format? The link is right here: https://docs.pylonsproject.org/projects/pastedeploy/en/latest/#config-format

Now the sections. vibe check time ✨

🌸 GLOBAL SECTION 🌸

This one is marked by the section named [DEFAULT]. Here are the parameters that are acceptable in this section.

  • bind_ip — the IP address the object server should bind to. Default is 0.0.0.0, which makes it bind to all available addresses. So it's like, “I'm everywhere, bestie” 🌈
  • bind_port — the TCP port the object server should bind to. Default is 6200.
  • keep_idle — value to set for socket TCP_KEEPIDLE. Default is 600.
  • bind_timeout — timeout to bind socket. Default is 30.
  • backlog — TCP backlog. Maximum number of allowed pending connections. Default is 4096.
  • workers — the number of pre-forked processes that will accept connections. Zero means no fork. Default is auto, which makes the server try to match the number of effective cpu cores if python multiprocessing is available (included with most python distributions >= 2.6), or fallback to one. Also worth noting: individual workers will use many eventlet co-routines to service multiple concurrent requests. slay 💅
  • max_clients — maximum number of clients one worker can process simultaneously (it will actually accept(2) N + 1). Setting this to one (1) will only handle one request at a time, without accepting another request concurrently. Default is 1024.
  • user — the system user that the object server will run as. Default is swift.
  • swift_dir — Swift configuration directory. Default is /etc/swift.
  • devices — parent directory of where devices are mounted. Default is /srv/node.
  • mount_check — whether or not to check if the devices are mounted to prevent accidentally writing to the root device. Default is set to true.
  • disable_fallocate — disable pre-allocate disk space for a file. Default is false.
  • servers_per_port — make object-server run this many worker processes per unique port of "local" ring devices across all storage policies. The default value of 0 disables this feature.
  • log_name — label used when logging. Default is swift.
  • log_facility — syslog log facility. Default is LOG_LOCAL0.
  • log_level — logging level. Default is INFO.
  • log_address — logging address. Default is /dev/log.
  • log_max_line_length — this caps the length of log lines to the value given; no limit if set to 0, which is the default.
  • log_custom_handlers — comma separated list of functions to call to setup custom log handlers. Functions get passed: conf, name, log_to_console, log_route, fmt, logger, adapted_logger. Default is empty.
  • log_udp_host — if set, log_udp_host will override log_address.
  • log_udp_port — UDP log port. Default is 514.
  • log_statsd_host — StatsD server. IPv4/IPv6 addresses and hostnames are supported. If a hostname resolves to an IPv4 and IPv6 address, the IPv4 address will be used.
  • log_statsd_port — default is 8125.
  • log_statsd_default_sample_rate — default is 1.
  • log_statsd_sample_rate_factor — default is 1.
  • log_statsd_metric_prefix — default is empty.
  • eventlet_debug — debug mode for eventlet library. Default is false.
  • fallocate_reserve — you can set fallocate_reserve to the number of bytes or percentage of disk space you'd like fallocate to reserve, whether there is space for the given file size or not. Percentage will be used if the value ends with a %. Default is 1%.
  • node_timeout — request timeout to external services. Default is 3 seconds.
  • conn_timeout — connection timeout to external services. Default is 0.5 seconds.
  • container_update_timeout — time to wait while sending a container update on object update. Default is 1 second.
  • client_timeout — time to wait while receiving each chunk of data from a client or another backend node. Default is 60.
  • network_chunk_size — default is 65536.
  • disk_chunk_size — default is 65536.
  • reclaim_age — time elapsed in seconds before an object can be reclaimed. Default is 604800 seconds.
  • commit_window — time in seconds during which a newly written non-durable data file will not be reclaimed. The value should be greater than zero and much less than reclaim_age. Default is 60.0 seconds.
  • nice_priority — modify scheduling priority of server processes. Niceness values range from -20 (most favorable to the process) to 19 (least favorable to the process). Default does not modify priority.
  • ionice_class — modify I/O scheduling class of server processes. I/O niceness class values are IOPRIO_CLASS_RT (realtime), IOPRIO_CLASS_BE (best-effort), and IOPRIO_CLASS_IDLE (idle). Default does not modify class and priority. Work only with ionice_priority.
  • ionice_priority — modify I/O scheduling priority of server processes. I/O niceness priority is a number which goes from 0 to 7. The higher the value, the lower the I/O priority of the process. Work only with ionice_class. Ignored if IOPRIO_CLASS_IDLE is set.

🦋 PIPELINE SECTION 🦋

This one is marked by section name [pipeline:main]. Here are the parameters acceptable in this section.

  • pipeline — used when you need to apply a number of filters. It's a list of filters ended by an application. The normal pipeline is "healthcheck recon object-server".

🌸 APP SECTION 🌸

This one shows up as the section named [app:object-server]. Here's everything you can drop in there, bestie 💖

  • use — the paste.deploy entry point for the object server. It's basically a pointer to the installed python egg. Normally it's egg:swift#object. 🥚
  • set log_name — the label it uses when logging. Default is object-server.
  • set log_facility — syslog log facility. Default is LOG_LOCAL0.
  • set log_level — logging level. Default is INFO.
  • set log_requests — turns request logging on. Default is True.
  • set log_address — logging address. Default is /dev/log.
  • max_upload_time — the default is 86400. ⏰
  • slow — the default is 0.
  • keep_cache_size — objects smaller than this don't get evicted from the buffercache once they've been read. Default is 5242880.
  • keep_cache_private — if true, objects for authenticated GET requests may be kept in buffer cache if they're small enough. Default is false.
  • mb_per_sync — on PUTs, sync data every n MB. Default is 512.
  • allowed_headers — a comma separated list of headers that can be set in metadata on an object. This list is on top of X-Object-Meta-* headers, and it cannot include Content-Type, etag, Content-Length, or deleted. The default is 'Content-Disposition, Content-Encoding, X-Delete-At, X-Object-Manifest, X-Static-Large-Object, Cache-Control, Content-Language, Expires, X-Robots-Tag'.
  • replication_server — the config parameter for building a specific server. To handle all verbs, including replication verbs, don't specify "replication_server" (that's the default). To only handle replication, set it to a True value (like "True" or "1"). To handle only non-replication verbs, set it to "False". Unless you have a separate replication network, you should not specify any value for "replication_server". 🫶
  • replication_concurrency — set this to restrict the number of concurrent incoming SSYNC requests. Set to 0 for unlimited (the default is 4). Heads up: SSYNC requests are only used by the object reconstructor or the object replicator when it's configured to use ssync.
  • replication_concurrency_per_device — set this to restrict the number of concurrent incoming SSYNC requests per device; set to 0 for unlimited requests per device. This can help control I/O to each device. It does not override replication_concurrency described above, so depending on your hardware or network capacity you may need to adjust both parameters. Defaults to 1.
  • replication_lock_timeout — number of seconds to wait for an existing replication device lock before giving up. Default is 15.
  • replication_failure_threshold
  • replication_failure_ratio — these two settings together control when the SSYNC subrequest handler will abort an incoming SSYNC attempt. An abort happens if there are at least threshold number of failures and the value of failures / successes exceeds the ratio. The defaults of 100 and 1.0 mean that at least 100 failures have to occur, and there have to be more failures than successes, for an abort to occur. 💀
  • splice — use splice() for zero-copy object GETs. This requires Linux kernel version 3.0 or greater. If you set "splice = yes" but the kernel doesn't support it, error messages will show up in the object server logs at startup, but your object servers should continue to function. Default is false.
  • node_timeout — request timeout to external services. Default is 3 seconds.
  • conn_timeout — connection timeout to external services. Default is 0.5 seconds.
  • container_update_timeout — time to wait while sending a container update on object update. Default is 1 second.
  • nice_priority — modify the scheduling priority of server processes. Niceness values range from -20 (most favorable to the process) to 19 (least favorable to the process). The default does not modify priority.
  • ionice_class — modify the I/O scheduling class of server processes. I/O niceness class values are IOPRIO_CLASS_RT (realtime), IOPRIO_CLASS_BE (best-effort) and IOPRIO_CLASS_IDLE (idle). The default does not modify class and priority. Only works together with ionice_priority.
  • ionice_priority — modify the I/O scheduling priority of server processes. I/O niceness priority is a number from 0 to 7. The higher the value, the lower the I/O priority of the process. Only works together with ionice_class. Ignored if IOPRIO_CLASS_IDLE is set.

🦋 FILTER SECTION 🦋

Any section whose name starts with "filter:" is a filter section, periodt. Filters are how you specify config parameters for specific swift middlewares. Here's what's available and what each one accepts, no gatekeeping 💅

[filter:healthcheck]

  • use — the paste.deploy entry point for the healthcheck middleware. This is the reference to the installed python egg. Normally it's egg:swift#healthcheck.
  • disable_path — an optional filesystem path which, if present, will make the healthcheck URL return "503 Service Unavailable" with a body of "DISABLED BY FILE".

[filter:recon]

  • use — the paste.deploy entry point for the recon middleware. This is the reference to the installed python egg. Normally it's egg:swift#recon.
  • recon_cache_path — this simply sets the directory where stats for a few items will be stored. Depending on how you deployed things, you may need to create this directory manually and make sure swift has read/write. The default is /var/cache/swift.
  • recon_lock_path — the default is /var/lock.

[filter:xprofile]

  • use — the paste.deploy entry point for the xprofile middleware. This is the reference to the installed python egg. Normally it's egg:swift#xprofile.
  • profile_module — this option lets you switch profilers, which should inherit from the python standard profiler. Currently the supported value can be 'cProfile', 'eventlet.green.profile' etc.
  • log_filename_prefix — this prefix gets combined with process ID and timestamp to name the profile data file. Make sure the executing user has permission to write into this path (missing path segments will be created, if necessary). If you enable profiling in more than one type of daemon, you must override it with a unique value — the default is /var/log/swift/profile/account.profile. 👑
  • dump_interval — the profile data gets dumped to local disk based on the above naming rule, in this interval. The default is 5.0.
  • dump_timestamp — be careful, bestie 😭 this option makes the profiler dump data into the file with a time stamp, which means there will be lots of files piled up in the directory. The default is false.
  • path — this is the path of the URL to access the mini web UI. The default is __profile__.
  • flush_at_shutdown — clear the data when the wsgi server shuts down. The default is false.
  • unwind — unwind the iterator of applications. Default is false.

✨ ADDITIONAL SECTIONS ✨

These next sections are used by the other swift-object services, like the replicator, updater, and auditor. 🧚

[object-replicator]

  • log_name — label used when logging. The default is object-replicator.
  • log_facility — syslog log facility. The default is LOG_LOCAL0.
  • log_level — logging level. The default is INFO.
  • log_address — logging address. The default is /dev/log.
  • daemonize — whether or not to run replication as a daemon. The default is yes.
  • run_pause [deprecated] — time in seconds to wait between replication passes. The default is 30.
  • interval — time in seconds to wait between replication passes. The default is 30.
  • concurrency — number of replication workers to spawn. The default is 1.
  • stats_interval — interval in seconds between logging replication statistics. The default is 300.
  • sync_method — the sync method to use; default is rsync but you can use ssync to try the EXPERIMENTAL all-swift-code-no-rsync-callouts method. Once ssync is verified as having performance comparable to, or better than, rsync, they plan to deprecate rsync so they can move on with more features for replication.
  • rsync_timeout — max duration of a partition rsync. The default is 900 seconds.
  • rsync_io_timeout — passed to rsync for I/O OP timeout. The default is 30 seconds.
  • rsync_compress — allow rsync to compress data which is transmitted to the destination node during sync. However, this is applicable only when the destination node is in a different region than the local one. NOTE: objects that are already compressed (for example: .tar.gz, .mp3) might slow down the syncing process. The default is false.
  • rsync_module — format of the rsync module where the replicator will send data. See etc/rsyncd.conf-sample for some usage examples. The default is empty.
  • node_timeout — request timeout to external services. The default is 10 seconds.
  • rsync_bwlimit — passed to rsync for bandwidth limit in kB/s. The default is 0 (unlimited).
  • http_timeout — max duration of an HTTP request. The default is 60 seconds.
  • lockup_timeout — attempts to kill all workers if nothing replicates for lockup_timeout seconds. The default is 1800 seconds.
  • ring_check_interval — the default is 15.
  • rsync_error_log_line_length — limits how long rsync error log lines are. 0 (default) means to log the entire line.
  • recon_cache_path — this simply sets the directory where stats for a few items will be stored. Depending on how you deployed things, you may need to create this directory manually and make sure swift has read/write. The default is /var/cache/swift. 🍓
  • handoffs_first — the flag that says "replicate handoffs before canonical partitions, bestie". It lets you force syncing and deleting handoffs fast. If you set it to a True value (like "True" or "1"), partitions that aren't supposed to be on the node get replicated first. The default is false.
  • handoff_delete — the number of replicas that are ensured in swift. If you set it to a number less than the number of replicas, object-replicator could delete local handoffs even if all replicas aren't ensured in the cluster 😭. Object-replicator would remove local handoff partition directories after syncing a partition when the number of successful responses is greater than or equal to this number. By default (auto), handoff partitions get removed when it has successfully replicated to all the canonical nodes.
okay listen up because this is important 🚨 handoffs_first and handoff_delete are options for a special case, like when the cluster's disk is full. These two options SHOULD NOT BE CHANGED, except in those extreme situations (e.g. disks filled up or are about to fill up). Anyway, DO NOT let your drives fill up. periodt.
  • nice_priority — modify the scheduling priority of server processes. Niceness values range from -20 (most favorable to the process) to 19 (least favorable to the process). The default does not modify priority.
  • ionice_class — modify the I/O scheduling class of server processes. I/O niceness class values are IOPRIO_CLASS_RT (realtime), IOPRIO_CLASS_BE (best-effort) and IOPRIO_CLASS_IDLE (idle). The default does not modify class and priority. Work only with ionice_priority.
  • ionice_priority — modify the I/O scheduling priority of server processes. I/O niceness priority is a number which goes from 0 to 7. The higher the value, the lower the I/O priority of the process. Work only with ionice_class. Ignored if IOPRIO_CLASS_IDLE is set.

🦋 [object-reconstructor] 🦋

  • log_name — label used when logging. The default is object-reconstructor.
  • log_facility — syslog log facility. The default is LOG_LOCAL0.
  • log_level — logging level. The default is INFO.
  • log_address — logging address. The default is /dev/log.
  • daemonize — whether or not to run replication as a daemon. The default is yes.
  • run_pause [deprecated] — time in seconds to wait between replication passes. The default is 30.
  • interval — time in seconds to wait between replication passes. The default is 30.
  • concurrency — number of replication workers to spawn. The default is 1.
  • stats_interval — interval in seconds between logging replication statistics. The default is 300.
  • node_timeout — request timeout to external services. The default is 10 seconds.
  • http_timeout — max duration of an HTTP request. The default is 60 seconds.
  • lockup_timeout — attempts to kill all workers if nothing replicates for lockup_timeout seconds. The default is 1800 seconds.
  • ring_check_interval — the default is 15.
  • recon_cache_path — the recon_cache_path simply sets the directory where stats for a few items will be stored. Depending on the method of deployment you may need to create this directory manually and ensure that swift has read/write. The default is /var/cache/swift.
  • handoffs_first — the flag to replicate handoffs prior to canonical partitions. It allows one to force syncing and deleting handoffs quickly. If set to a True value (e.g. "True" or "1"), partitions that are not supposed to be on the node will be replicated first. The default is false.

🍓 [object-updater] 🍓

  • log_name — label used when logging. The default is object-updater.
  • log_facility — syslog log facility. The default is LOG_LOCAL0.
  • log_level — logging level. The default is INFO.
  • log_address — logging address. The default is /dev/log.
  • interval — minimum time for a pass to take. The default is 300 seconds.
  • concurrency — number of updater workers to spawn. The default is 1.
  • node_timeout — request timeout to external services. The default is 10 seconds.
  • objects_per_second — maximum objects updated per second. Should be tuned according to individual system specs. 0 is unlimited. The default is 50.
  • slowdown [deprecated] — slowdown will sleep that amount between objects. The default is 0.01 seconds. Deprecated in favor of objects_per_second.
  • recon_cache_path — the recon_cache_path simply sets the directory where stats for a few items will be stored. Depending on the method of deployment you may need to create this directory manually and ensure that swift has read/write. The default is /var/cache/swift.
  • nice_priority — modify the scheduling priority of server processes. Niceness values range from -20 (most favorable to the process) to 19 (least favorable to the process). The default does not modify priority.
  • ionice_class — modify the I/O scheduling class of server processes. I/O niceness class values are IOPRIO_CLASS_RT (realtime), IOPRIO_CLASS_BE (best-effort) and IOPRIO_CLASS_IDLE (idle). The default does not modify class and priority. Work only with ionice_priority.
  • ionice_priority — modify the I/O scheduling priority of server processes. I/O niceness priority is a number which goes from 0 to 7. The higher the value, the lower the I/O priority of the process. Work only with ionice_class. Ignored if IOPRIO_CLASS_IDLE is set.

💖 [object-auditor] 💖

  • log_name — label used when logging. The default is object-auditor.
  • log_facility — syslog log facility. The default is LOG_LOCAL0.
  • log_level — logging level. The default is INFO.
  • log_address — logging address. The default is /dev/log.
  • disk_chunk_size — the default is 65536.
  • files_per_second — maximum files audited per second. Should be tuned according to individual system specs. 0 is unlimited. The default is 20.
  • bytes_per_second — maximum bytes audited per second. Should be tuned according to individual system specs. 0 is unlimited. The default is 10000000.
  • concurrency — number of auditor workers to spawn. The default is 1.
  • log_time — the default is 3600 seconds.
  • zero_byte_files_per_second — the default is 50.
  • recon_cache_path — the recon_cache_path simply sets the directory where stats for a few items will be stored. Depending on the method of deployment you may need to create this directory manually and ensure that swift has read/write. The default is /var/cache/swift.
  • object_size_stats — takes a comma separated list of ints. If set, the object auditor will increment a counter for every object whose size is <= to the given break points and report the result after a full scan.
  • rsync_tempfile_timeout — time elapsed in seconds before rsync tempfiles will be unlinked. Config value of "auto" will try to use object-replicator's rsync_timeout + 900 or fall-back to 86400 (1 day).
  • nice_priority — modify the scheduling priority of server processes. Niceness values range from -20 (most favorable to the process) to 19 (least favorable to the process). The default does not modify priority.
  • ionice_class — modify the I/O scheduling class of server processes. I/O niceness class values are IOPRIO_CLASS_RT (realtime), IOPRIO_CLASS_BE (best-effort) and IOPRIO_CLASS_IDLE (idle). The default does not modify class and priority. Work only with ionice_priority.
  • ionice_priority — modify the I/O scheduling priority of server processes. I/O niceness priority is a number which goes from 0 to 7. The higher the value, the lower the I/O priority of the process. Work only with ionice_class. Ignored if IOPRIO_CLASS_IDLE is set.

👑 [object-expirer] 👑

  • log_name — label used when logging. The default is object-expirer.
  • log_facility — syslog log facility. The default is LOG_LOCAL0.
  • log_level — logging level. The default is INFO.
  • log_address — logging address. The default is /dev/log.
  • interval — minimum time for a pass to take. The default is 300 seconds.
  • report_interval — minimum time for a pass to report. The default is 300 seconds.
  • request_tries — the number of times the expirer's internal client will attempt any given request in the event of failure. The default is 3.
  • concurrency — number of expirer workers to spawn. The default is 1.
  • dequeue_from_legacy — the flag to execute legacy style expirer tasks. The default is false.
  • processes — processes can only be used in conjunction with dequeue_from_legacy. Processes is how many parts to divide the legacy work into, one part per process that will be doing the work. Processes set 0 means that a single process will be doing all the legacy work. Processes can also be specified on the command line and will override the config value. The default is 0.
  • process — process can only be used in conjunction with dequeue_from_legacy. Process is which of the parts a particular legacy process will work on, process can also be specified on the command line and will override the config value, process is "zero based", if you want to use 3 processes, you should run processes with process set to 0, 1, and 2. The default is 0.
  • reclaim_age — okay so if the source object isn't there, the expirer will keep re-attempting the expire for up to reclaim_age seconds before it fully gives up and yeets the task out of the queue 💀. Default is 604800 seconds (= 1 week).
  • delay_reaping_account/container — so normally the expirer starts reaping expired objects immediately, no chill. But bestie, you can add options prefixed with delay_reaping_ shaped like delay_reaping_<ACCT>[/<CNTR>] and the expirer will delay processing tasks in that account or account/container 🫶. The <ACCT>[/<CNTR>] part of the config option names should url-quote the paths (yes fr fr). The value is in seconds. Default is no delay for any tasks.
  • round_robin_task_cache_size — the number of task objects to cache before processing. simple as that 💅.
  • recon_cache_path — path to the recon cache directory. Default is /var/cache/swift ✨.
  • nice_priority — lets you modify the scheduling priority of server processes. Niceness values range from -20 (most favorable to the process) to 19 (least favorable to the process). The default does not modify priority.
  • ionice_class — modify the I/O scheduling class of server processes. I/O niceness class values are IOPRIO_CLASS_RT (realtime), IOPRIO_CLASS_BE (best-effort) and IOPRIO_CLASS_IDLE (idle). The default does not modify class and priority. It only works with ionice_priority 👑.
  • ionice_priority — modify the I/O scheduling priority of server processes. I/O niceness priority is a number which goes from 0 to 7. The higher the value, the lower the I/O priority of the process. Works only with ionice_class. It's ignored if IOPRIO_CLASS_IDLE is set, so don't even try it 🙈.

📚 DOCUMENTATION 📚

More in-depth documentation about the swift-object-server and also about OpenStack Swift as a whole can be found at:

like, go read it, it's giving academic but in a good way 🦋

👀 SEE ALSO 👀

and that's a wrap on object-server.conf(5), bestie 🎀 this page is from the Linux manual, dated 8/26/2011 — yes it's an elder but she still ate 💅 go forth and configure responsibly, mwah 💖✨🌸

📦 where this page comes from