[ceph.git] / ceph / src / test / cli / crushtool / help.t

  $ crushtool --help
  usage: crushtool ...
  
  Display, modify and test a crush map
  
  There are five stages, running one after the other:
  
   - input/build
   - tunables adjustments
   - modifications
   - display/test
   - output
  
  Options that are not specific to a stage.
  
     [--infn|-i infile]
                           read the crush map from infile
  
  Options for the input/build stage
  
     --decompile|-d map    decompile a crush map to source
     [--outfn|-o outfile]
                           specify output for for (de)compilation
     --compile|-c map.txt  compile a map from source
     --enable-unsafe-tunables
                           compile with unsafe tunables
     --build --num_osds N layer1 ...
                           build a new map, where each 'layer' is
                           'name (uniform|straw2|straw|list|tree) size'
  
  Options for the tunables adjustments stage
  
     --set-choose-local-tries N
                           set choose local retries before re-descent
     --set-choose-local-fallback-tries N
                           set choose local retries using fallback
                           permutation before re-descent
     --set-choose-total-tries N
                           set choose total descent attempts
     --set-chooseleaf-descend-once <0|1>
                           set chooseleaf to (not) retry the recursive descent
     --set-chooseleaf-vary-r <0|1>
                           set chooseleaf to (not) vary r based on parent
     --set-chooseleaf-stable <0|1>
                           set chooseleaf firstn to (not) return stable results
  
  Options for the modifications stage
  
     -i mapfn --add-item id weight name [--loc type name ...]
                           insert an item into the hierarchy at the
                           given location
     -i mapfn --update-item id weight name [--loc type name ...]
                           insert or move an item into the hierarchy at the
                           given location
     -i mapfn --remove-item name
                           remove the given item
     -i mapfn --reweight-item name weight
                           reweight a given item (and adjust ancestor
                           weights as needed)
     -i mapfn --reweight   recalculate all bucket weights
  
  Options for the display/test stage
  
     -f --format           the format of --dump, defaults to json-pretty
                           can be one of json, json-pretty, xml, xml-pretty,
                           table, table-kv, html, html-pretty
     --dump                dump the crush map
     --tree                print map summary as a tree
     --check [max_id]      check if any item is referencing an unknown name/type
     -i mapfn --show-location id
                           show location for given device id
     -i mapfn --test       test a range of inputs on the map
        [--min-x x] [--max-x x] [--x x]
        [--min-rule r] [--max-rule r] [--rule r] [--ruleset rs]
        [--num-rep n]
        [--pool-id n]      specifies pool id
        [--batches b]      split the CRUSH mapping into b > 1 rounds
        [--weight|-w devno weight]
                           where weight is 0 to 1.0
        [--simulate]       simulate placements using a random
                           number generator in place of the CRUSH
                           algorithm
     --show-utilization    show OSD usage
     --show-utilization-all
                           include zero weight items
     --show-statistics     show chi squared statistics
     --show-mappings       show mappings
     --show-bad-mappings   show bad mappings
     --show-choose-tries   show choose tries histogram
     --output-name name
                           prepend the data file(s) generated during the
                           testing routine with name
     --output-csv
                           export select data generated during testing routine
                           to CSV files for off-line post-processing
                           use --help-output for more information
  
  Options for the output stage
  
     [--outfn|-o outfile]
                           specify output for modified crush map
  
  $ crushtool --help-output
  data output from testing routine ...
             absolute_weights
                    the decimal weight of each OSD
                    data layout: ROW MAJOR
                                 OSD id (int), weight (int)
             batch_device_expected_utilization_all
                    the expected number of objects each OSD should receive per placement batch
                    which may be a decimal value
                    data layout: COLUMN MAJOR
                                 round (int), objects expected on OSD 0...OSD n (float)
             batch_device_utilization_all
                    the number of objects stored on each OSD during each placement round
                    data layout: COLUMN MAJOR
                                 round (int), objects stored on OSD 0...OSD n (int)
             device_utilization_all
                    the number of objects stored on each OSD at the end of placements
                    data_layout: ROW MAJOR
                                 OSD id (int), objects stored (int), objects expected (float)
             device_utilization
                    the number of objects stored on each OSD marked 'up' at the end of placements
                    data_layout: ROW MAJOR
                                 OSD id (int), objects stored (int), objects expected (float)
             placement_information
                    the map of input -> OSD
                    data_layout: ROW MAJOR
                                 input (int), OSD's mapped (int)
             proportional_weights_all
                    the proportional weight of each OSD specified in the CRUSH map
                    data_layout: ROW MAJOR
                                 OSD id (int), proportional weight (float)
             proportional_weights
                    the proportional weight of each 'up' OSD specified in the CRUSH map
                    data_layout: ROW MAJOR
                                 OSD id (int), proportional weight (float)
Commit	Line	Data
7c673cae FG	1	$ crushtool --help
	2	usage: crushtool ...
	3
	4	Display, modify and test a crush map
	5
	6	There are five stages, running one after the other:
	7
	8	- input/build
	9	- tunables adjustments
	10	- modifications
	11	- display/test
	12	- output
	13
	14	Options that are not specific to a stage.
	15
	16	[--infn\|-i infile]
	17	read the crush map from infile
	18
	19	Options for the input/build stage
	20
	21	--decompile\|-d map decompile a crush map to source
	22	[--outfn\|-o outfile]
	23	specify output for for (de)compilation
	24	--compile\|-c map.txt compile a map from source
	25	--enable-unsafe-tunables
	26	compile with unsafe tunables
	27	--build --num_osds N layer1 ...
	28	build a new map, where each 'layer' is
	29	'name (uniform\|straw2\|straw\|list\|tree) size'
	30
	31	Options for the tunables adjustments stage
	32
	33	--set-choose-local-tries N
	34	set choose local retries before re-descent
	35	--set-choose-local-fallback-tries N
	36	set choose local retries using fallback
	37	permutation before re-descent
	38	--set-choose-total-tries N
	39	set choose total descent attempts
	40	--set-chooseleaf-descend-once <0\|1>
	41	set chooseleaf to (not) retry the recursive descent
	42	--set-chooseleaf-vary-r <0\|1>
	43	set chooseleaf to (not) vary r based on parent
	44	--set-chooseleaf-stable <0\|1>
	45	set chooseleaf firstn to (not) return stable results
	46
	47	Options for the modifications stage
	48
	49	-i mapfn --add-item id weight name [--loc type name ...]
	50	insert an item into the hierarchy at the
	51	given location
	52	-i mapfn --update-item id weight name [--loc type name ...]
	53	insert or move an item into the hierarchy at the
	54	given location
	55	-i mapfn --remove-item name
	56	remove the given item
	57	-i mapfn --reweight-item name weight
	58	reweight a given item (and adjust ancestor
	59	weights as needed)
	60	-i mapfn --reweight recalculate all bucket weights
	61
	62	Options for the display/test stage
	63
	64	-f --format the format of --dump, defaults to json-pretty
65	can be one of json, json-pretty, xml, xml-pretty,
66	table, table-kv, html, html-pretty
67	--dump dump the crush map
68	--tree print map summary as a tree
69	--check [max_id] check if any item is referencing an unknown name/type
70	-i mapfn --show-location id
71	show location for given device id
72	-i mapfn --test test a range of inputs on the map
73	[--min-x x] [--max-x x] [--x x]
74	[--min-rule r] [--max-rule r] [--rule r] [--ruleset rs]
75	[--num-rep n]
76	[--pool-id n] specifies pool id
77	[--batches b] split the CRUSH mapping into b > 1 rounds
78	[--weight\|-w devno weight]
79	where weight is 0 to 1.0
80	[--simulate] simulate placements using a random
81	number generator in place of the CRUSH
82	algorithm
83	--show-utilization show OSD usage
84	--show-utilization-all
85	include zero weight items
86	--show-statistics show chi squared statistics
87	--show-mappings show mappings
88	--show-bad-mappings show bad mappings
89	--show-choose-tries show choose tries histogram
90	--output-name name
91	prepend the data file(s) generated during the
92	testing routine with name
93	--output-csv
94	export select data generated during testing routine
95	to CSV files for off-line post-processing
96	use --help-output for more information
97
98	Options for the output stage
99
100	[--outfn\|-o outfile]
101	specify output for modified crush map
102
103	$ crushtool --help-output
104	data output from testing routine ...
105	absolute_weights
106	the decimal weight of each OSD
107	data layout: ROW MAJOR
108	OSD id (int), weight (int)
109	batch_device_expected_utilization_all
110	the expected number of objects each OSD should receive per placement batch
111	which may be a decimal value
112	data layout: COLUMN MAJOR
113	round (int), objects expected on OSD 0...OSD n (float)
114	batch_device_utilization_all
115	the number of objects stored on each OSD during each placement round
116	data layout: COLUMN MAJOR
117	round (int), objects stored on OSD 0...OSD n (int)
118	device_utilization_all
119	the number of objects stored on each OSD at the end of placements
120	data_layout: ROW MAJOR
121	OSD id (int), objects stored (int), objects expected (float)
122	device_utilization
123	the number of objects stored on each OSD marked 'up' at the end of placements
124	data_layout: ROW MAJOR
125	OSD id (int), objects stored (int), objects expected (float)
126	placement_information
127	the map of input -> OSD
128	data_layout: ROW MAJOR
129	input (int), OSD's mapped (int)
130	proportional_weights_all
131	the proportional weight of each OSD specified in the CRUSH map
132	data_layout: ROW MAJOR
133	OSD id (int), proportional weight (float)
134	proportional_weights
135	the proportional weight of each 'up' OSD specified in the CRUSH map
136	data_layout: ROW MAJOR
137	OSD id (int), proportional weight (float)