328
328
bind_ip 0.0.0.0 IP Address for server to bind to
329
329
bind_port 6000 Port for server to bind to
330
330
bind_timeout 30 Seconds to attempt bind before giving up
331
workers 1 Number of workers to fork
331
workers auto Override the number of pre-forked workers
332
that will accept connections. If set it
333
should be an integer, zero means no fork. If
334
unset, it will try to default to the number
335
of effective cpu cores and fallback to one.
336
Increasing the number of workers may reduce
337
the possibility of slow file system
338
operations in one request from negatively
339
impacting other requests, but may not be as
340
efficient as tuning :ref:`threads_per_disk
341
<object-server-options>`
332
342
max_clients 1024 Maximum number of clients one worker can
333
343
process simultaneously (it will actually
334
344
accept(2) N + 1). Setting this to one (1)
335
345
will only handle one request at a time,
336
346
without accepting another request
337
concurrently. By increasing the number of
338
workers to a much higher value, one can
339
reduce the impact of slow file system
340
operations in one request from negatively
341
impacting other requests.
342
348
disable_fallocate false Disable "fast fail" fallocate checks if the
343
349
underlying filesystem does not support it.
344
350
log_custom_handlers None Comma-separated list of functions to call
462
470
bind_ip 0.0.0.0 IP Address for server to bind to
463
471
bind_port 6001 Port for server to bind to
464
472
bind_timeout 30 Seconds to attempt bind before giving up
465
workers 1 Number of workers to fork
473
workers auto Override the number of pre-forked workers
474
that will accept connections. If set it
475
should be an integer, zero means no fork. If
476
unset, it will try to default to the number
477
of effective cpu cores and fallback to one.
478
Increasing the number of workers may reduce
479
the possibility of slow file system
480
operations in one request from negatively
481
impacting other requests. See
482
:ref:`general-service-tuning`
466
483
max_clients 1024 Maximum number of clients one worker can
467
484
process simultaneously (it will actually
468
485
accept(2) N + 1). Setting this to one (1)
469
486
will only handle one request at a time,
470
487
without accepting another request
471
concurrently. By increasing the number of
472
workers to a much higher value, one can
473
reduce the impact of slow file system
474
operations in one request from negatively
475
impacting other requests.
476
489
user swift User to run as
477
490
disable_fallocate false Disable "fast fail" fallocate checks if the
478
491
underlying filesystem does not support it.
582
595
bind_ip 0.0.0.0 IP Address for server to bind to
583
596
bind_port 6002 Port for server to bind to
584
597
bind_timeout 30 Seconds to attempt bind before giving up
585
workers 1 Number of workers to fork
598
workers auto Override the number of pre-forked workers
599
that will accept connections. If set it
600
should be an integer, zero means no fork. If
601
unset, it will try to default to the number
602
of effective cpu cores and fallback to one.
603
Increasing the number of workers may reduce
604
the possibility of slow file system
605
operations in one request from negatively
606
impacting other requests. See
607
:ref:`general-service-tuning`
586
608
max_clients 1024 Maximum number of clients one worker can
587
609
process simultaneously (it will actually
588
610
accept(2) N + 1). Setting this to one (1)
589
611
will only handle one request at a time,
590
612
without accepting another request
591
concurrently. By increasing the number of
592
workers to a much higher value, one can
593
reduce the impact of slow file system
594
operations in one request from negatively
595
impacting other requests.
596
614
user swift User to run as
597
615
db_preallocation off If you don't mind the extra disk space usage in
598
616
overhead, you can turn this on to preallocate
696
714
bind_timeout 30 Seconds to attempt bind before
698
716
swift_dir /etc/swift Swift configuration directory
699
workers 1 Number of workers to fork
717
workers auto Override the number of
718
pre-forked workers that will
719
accept connections. If set it
720
should be an integer, zero
721
means no fork. If unset, it
722
will try to default to the
723
number of effective cpu cores
724
and fallback to one. See
725
:ref:`general-service-tuning`
700
726
max_clients 1024 Maximum number of clients one
701
727
worker can process
702
728
simultaneously (it will
705
731
will only handle one request at
706
732
a time, without accepting
708
concurrently. By increasing the
709
number of workers to a much
710
higher value, one can reduce
711
the impact of slow file system
712
operations in one request from
713
negatively impacting other
715
735
user swift User to run as
716
736
cert_file Path to the ssl .crt. This
717
737
should be enabled for testing
968
990
the best all-around choice. If you decide to use a filesystem other than
969
991
XFS, we highly recommend thorough testing.
971
If you are using XFS, some settings that can dramatically impact
972
performance. We recommend the following when creating the XFS
975
mkfs.xfs -i size=1024 -f /dev/sda1
993
For distros with more recent kernels (for example Ubuntu 12.04 Precise),
994
we recommend using the default settings (including the default inode size
995
of 256 bytes) when creating the file system::
999
In the last couple of years, XFS has made great improvements in how inodes
1000
are allocated and used. Using the default inode size no longer has an
1001
impact on performance.
1003
For distros with older kernels (for example Ubuntu 10.04 Lucid),
1004
some settings can dramatically impact performance. We recommend the
1005
following when creating the file system::
1007
mkfs.xfs -i size=1024 /dev/sda1
977
1009
Setting the inode size is important, as XFS stores xattr data in the inode.
978
1010
If the metadata is too large to fit in the inode, a new extent is created,
979
1011
which can cause quite a performance problem. Upping the inode size to 1024
980
1012
bytes provides enough room to write the default metadata, plus a little
981
headroom. We do not recommend running Swift on RAID, but if you are using
1015
The following example mount options are recommended when using XFS::
1017
mount -t xfs -o noatime,nodiratime,nobarrier,logbufs=8 /dev/sda1 /srv/node/sda
1019
We do not recommend running Swift on RAID, but if you are using
982
1020
RAID it is also important to make sure that the proper sunit and swidth
983
1021
settings get set so that XFS can make most efficient use of the RAID array.
985
We also recommend the following example mount options when using XFS::
987
mount -t xfs -o noatime,nodiratime,nobarrier,logbufs=8 /dev/sda1 /srv/node/sda
989
1023
For a standard swift install, all data drives are mounted directly under
990
1024
/srv/node (as can be seen in the above example of mounting /def/sda1 as
991
1025
/srv/node/sda). If you choose to mount the drives in another directory,