Lines Matching +full:asi +full:- +full:format
4 opensm \- InfiniBand subnet manager and administration (SM/SA)
8 [\-\-version]]
9 [\-F | \-\-config <file_name>]
10 [\-c(reate-config) <file_name>]
11 [\-g(uid) <GUID in hex>]
12 [\-l(mc) <LMC>]
13 [\-p(riority) <PRIORITY>]
14 [\-\-smkey <SM_Key>]
15 [\-\-sm_sl <SL number>]
16 [\-r(eassign_lids)]
17 [\-R <engine name(s)> | \-\-routing_engine <engine name(s)>]
18 [\-\-do_mesh_analysis]
19 [\-\-lash_start_vl <vl number>]
20 [\-A | \-\-ucast_cache]
21 [\-z | \-\-connect_roots]
22 [\-M <file name> | \-\-lid_matrix_file <file name>]
23 [\-U <file name> | \-\-lfts_file <file name>]
24 [\-S | \-\-sadb_file <file name>]
25 [\-a | \-\-root_guid_file <path to file>]
26 [\-u | \-\-cn_guid_file <path to file>]
27 [\-G | \-\-io_guid_file <path to file>]
28 [\-\-port\-shifting]
29 [\-\-scatter\-ports <random seed>]
30 [\-H | \-\-max_reverse_hops <max reverse hops allowed>]
31 [\-X | \-\-guid_routing_order_file <path to file>]
32 [\-m | \-\-ids_guid_file <path to file>]
33 [\-o(nce)]
34 [\-s(weep) <interval>]
35 [\-t(imeout) <milliseconds>]
36 [\-\-retries <number>]
37 [\-\-maxsmps <number>]
38 [\-\-console [off | local | socket | loopback]]
39 [\-\-console-port <port>]
40 [\-i | \-\-ignore_guids <equalize-ignore-guids-file>]
41 [\-w | \-\-hop_weights_file <path to file>]
42 [\-O | \-\-port_search_ordering_file <path to file>]
43 [\-O | \-\-dimn_ports_file <path to file>] (DEPRECATED)
44 [\-f <log file path> | \-\-log_file <log file path> ]
45 [\-L | \-\-log_limit <size in MB>] [\-e(rase_log_file)]
46 [\-P(config) <partition config file> ]
47 [\-N | \-\-no_part_enforce] (DEPRECATED)
48 [\-Z | \-\-part_enforce [both | in | out | off]]
49 [\-W | \-\-allow_both_pkeys]
50 [\-Q | \-\-qos [\-Y | \-\-qos_policy_file <file name>]]
51 [\-\-congestion\-control]
52 [\-\-cckey <key>]
53 [\-y | \-\-stay_on_fatal]
54 [\-B | \-\-daemon]
55 [\-J | \-\-pidfile <file_name>]
56 [\-I | \-\-inactive]
57 [\-\-perfmgr]
58 [\-\-perfmgr_sweep_time_s <seconds>]
59 [\-\-prefix_routes_file <path>]
60 [\-\-consolidate_ipv6_snm_req]
61 [\-\-log_prefix <prefix text>]
62 [\-\-torus_config <path to file>]
63 [\-v(erbose)] [\-V] [\-D <flags>] [\-d(ebug) <number>]
64 [\-h(elp)] [\-?]
94 (Note that when a fatal and non-recoverable error occurs, opensm will exit.)
102 \fB\-\-version\fR
105 \fB\-F\fR, \fB\-\-config\fR <config file>
109 \fB\-c\fR, \fB\-\-create-config\fR <file name>
113 \fB\-g\fR, \fB\-\-guid\fR <GUID in hex>
119 Without -g, OpenSM tries to use the default port.
121 \fB\-l\fR, \fB\-\-lmc\fR <LMC value>
124 The LMC value must be in the range 0-7.
129 Without -l, OpenSM defaults to LMC = 0, which allows
132 \fB\-p\fR, \fB\-\-priority\fR <Priority value>
138 \fB\-\-smkey\fR <SM_Key value>
145 \fB\-\-sm_sl\fR <SL number>
149 \fB\-r\fR, \fB\-\-reassign_lids\fR
151 end nodes. Specifying -r on a running subnet
153 Without -r, OpenSM attempts to preserve existing
156 \fB\-R\fR, \fB\-\-routing_engine\fR <Routing engine names>
163 Supported engines: minhop, updn, dnup, file, ftree, lash, dor, torus-2QoS,
166 \fB\-\-do_mesh_analysis\fR
171 \fB\-\-lash_start_vl\fR <vl number>
175 \fB\-A\fR, \fB\-\-ucast_cache\fR
187 \fB\-z\fR, \fB\-\-connect_roots\fR
189 fat-tree) to make connectivity between root switches and in
193 \fB\-M\fR, \fB\-\-lid_matrix_file\fR <file name>
198 \fB\-U\fR, \fB\-\-lfts_file\fR <file name>
203 \fB\-S\fR, \fB\-\-sadb_file\fR <file name>
207 \fB\-a\fR, \fB\-\-root_guid_file\fR <file name>
208 Set the root nodes for the Up/Down or Fat-Tree routing
211 \fB\-u\fR, \fB\-\-cn_guid_file\fR <file name>
212 Set the compute nodes for the Fat-Tree or DFSSSP/SSSP routing algorithms
215 \fB\-G\fR, \fB\-\-io_guid_file\fR <file name>
216 Set the I/O nodes for the Fat-Tree or DFSSSP/SSSP routing algorithms
219 In the case of Fat-Tree routing:
221 I/O nodes are non-CN nodes allowed to use up to max_reverse_hops switches
230 \fB\-\-port\-shifting\fR
237 \fB\-\-scatter\-ports\fR <random seed>
239 using a round-robin algorithm (which is the default). Value supplied
243 \fB\-H\fR, \fB\-\-max_reverse_hops\fR <max reverse hops allowed>
247 \fB\-m\fR, \fB\-\-ids_guid_file\fR <file name>
252 \fB\-X\fR, \fB\-\-guid_routing_order_file\fR <file name>
257 \fB\-o\fR, \fB\-\-once\fR
261 \fB\-s\fR, \fB\-\-sweep\fR <interval value>
263 subnet sweeps. Specifying -s 0 disables sweeping.
264 Without -s, OpenSM defaults to a sweep interval of
267 \fB\-t\fR, \fB\-\-timeout\fR <value>
271 Without -t, OpenSM defaults to a timeout value of
274 \fB\-\-retries\fR <number>
277 Without --retries, OpenSM defaults to 3 retries
280 \fB\-\-maxsmps\fR <number>
283 Specifying \-\-maxsmps 0 allows unlimited outstanding
285 Without \-\-maxsmps, OpenSM defaults to a maximum of
288 \fB\-\-console [off | local | loopback | socket]\fR
293 will only be available if OpenSM was built with --enable-console-loopback
294 (default yes) and --enable-console-socket (default no) respectively.
296 \fB\-\-console-port\fR <port>
299 --enable-console-socket.
301 \fB\-i\fR, \fB\-\-ignore_guids\fR <equalize-ignore-guids-file>
306 \fB\-w\fR, \fB\-\-hop_weights_file\fR <path to file>
315 \fB\-O\fR, \fB\-\-port_search_ordering_file\fR <path to file>
321 number, with leading 0x) followed by a list of non-zero port numbers,
330 hex number, with leading 0x) followed by a list of non-zero port numbers,
336 \fB\-O\fR, \fB\-\-dimn_ports_file\fR <path to file> \fB(DEPRECATED)\fR
337 This is a deprecated flag. Please use \fB\-\-port_search_ordering_file\fR instead.
341 number, with leading 0x) followed by a list of non-zero port numbers,
347 \fB\-x\fR, \fB\-\-honor_guid2lid\fR
353 \fB\-f\fR, \fB\-\-log_file\fR <file name>
356 For the log to go to standard output use -f stdout.
358 \fB\-L\fR, \fB\-\-log_limit\fR <size in MB>
363 \fB\-e\fR, \fB\-\-erase_log_file\fR
368 \fB\-P\fR, \fB\-\-Pconfig\fR <partition config file>
372 \fB\-\-prefix_routes_file\fR <file name>
374 off-subnet DGIDs. By default, the SA fails such queries. The
376 section below describes the format of the configuration file.
377 The default path is \fB\%/etc/opensm/prefix\-routes.conf\fP.
379 \fB\-Q\fR, \fB\-\-qos\fR
382 \fB\-Y\fR, \fB\-\-qos_policy_file\fR <file name>
384 name is \fB\%/etc/opensm/qos-policy.conf\fP. See
388 \fB\-\-congestion_control\fR
392 \fB\-\-cc_key\fR <key>
397 \fB\-N\fR, \fB\-\-no_part_enforce\fR \fB(DEPRECATED)\fR
398 This is a deprecated flag. Please use \fB\-\-part_enforce\fR instead.
401 \fB\-Z\fR, \fB\-\-part_enforce\fR [both | in | out | off]
406 \fB\-W\fR, \fB\-\-allow_both_pkeys\fR
411 \fB\-y\fR, \fB\-\-stay_on_fatal\fR
417 \fB\-B\fR, \fB\-\-daemon\fR
418 Run in daemon mode - OpenSM will run in the background.
420 \fB\-J\fR, \fB\-\-pidfile <file_name>\fR
424 \fB\-I\fR, \fB\-\-inactive\fR
430 \fB\-\-perfmgr\fR
431 Enable the perfmgr. Only takes effect if --enable-perfmgr was specified at
432 configure time. See performance-manager-HOWTO.txt in opensm doc for
435 \fB\-\-perfmgr_sweep_time_s\fR <seconds>
438 effect if --enable-perfmgr was specified at configure time.
440 .BI --consolidate_ipv6_snm_req
444 \fB\-\-log_prefix\fR <prefix text>
448 fabrics. For example, in a dual-fabric (or dual-rail) IB cluster, the prefix
451 \fB\-\-torus_config\fR <path to torus\-2QoS config file>
453 information needed for the torus-2QoS routing engine. The default
454 name is \fB\%/etc/opensm/torus-2QoS.conf\fP
456 \fB\-v\fR, \fB\-\-verbose\fR
458 The -v option may be specified multiple times
460 See the -D option for more information about
463 \fB\-V\fR
466 The -V option is equivalent to \'-D 0xFF -d 2\'.
467 See the -D option for more information about
470 \fB\-D\fR <value>
472 A flags field must follow the -D option.
477 ---- -----------------
478 0x01 - ERROR (error messages)
479 0x02 - INFO (basic messages, low volume)
480 0x04 - VERBOSE (interesting stuff, moderate volume)
481 0x08 - DEBUG (diagnostic, high volume)
482 0x10 - FUNCS (function entry/exit, very high volume)
483 0x20 - FRAMES (dumps all SMP and GMP frames)
484 0x40 - ROUTING (dump FDB routing information)
485 0x80 - SYS (syslog at LOG_INFO level in addition to OpenSM logging)
487 Without -D, OpenSM defaults to ERROR + INFO (0x3).
488 Specifying -D 0 disables all messages.
489 Specifying -D 0xFF enables all messages (see -V).
491 the transaction timeout with the -t option.
493 \fB\-d\fR, \fB\-\-debug\fR <value>
496 The number following -d selects the debug
500 --- -----------------
501 -d0 - Ignore other SM nodes
502 -d1 - Force single threaded dispatching
503 -d2 - Force log flushing after each log message
504 -d3 - Disable multicast support
506 \fB\-h\fR, \fB\-\-help\fR
509 \fB\-?\fR
516 OSM_TMP_DIR - controls the directory in which the temporary files generated by
517 opensm are created. These files are: opensm-subnet.lst, opensm.fdbs, and
520 OSM_CACHE_DIR - opensm stores certain data to the disk such that subsequent
524 guid2lid - stores the LID range assigned to each GUID
525 guid2mkey - stores the MKey previously assiged to each GUID
526 neighbors - stores a map of the GUIDs at either end of each link
540 by using the --Pconfig (-P) option with OpenSM.
563 File Format
570 General file format:
577 PartitionName - string, will be used with logging. When
579 PKey - P_Key value for this partition. Only low 15
582 indx0 - indicates that this pkey should be inserted in
584 ipoib_bc_flags - used to indicate/specify IPoIB capability of
587 defmember=full|limited|both - specifies default membership for
594 ipoib - indicates that this partition may be used for
608 PortGUID - GUID of partition member EndPort.
611 full, limited, - indicates full and/or limited membership for
620 - gid specified is verified to be a Multicast
630 rate=<val> - specifies rate for this MC group
632 mtu=<val> - specifies MTU for this MC group
634 sl=<val> - specifies SL for this MC group
636 scope=<val> - specifies scope for this MC group
643 Q_Key=<val> - specifies the Q_Key for this MC group
649 TClass=<val> - specifies tclass for this MC group
651 FlowLabel=<val> - specifies FlowLabel for this MC group
660 - 'ALL' means all end ports in this subnet.
661 - 'ALL_CAS' means all Channel Adapter end ports in this subnet.
662 - 'ALL_SWITCHES' means all Switch end ports in this subnet.
663 - 'ALL_ROUTERS' means all Router end ports in this subnet.
664 - 'SELF' means subnet manager's port.
701 mgid=ff12:601b::16 # MLDv2-capable routers
717 There are a set of QoS related low-level configuration parameters.
721 qos_max_vls - The maximum number of VLs that will be on the subnet
722 qos_high_limit - The limit of High Priority component of VL
724 qos_vlarb_low - Low priority VL Arbitration table (IBA 7.6.9)
726 qos_vlarb_high - High priority VL Arbitration table (IBA 7.6.9)
730 qos_sl2vl - SL2VL Mapping table (IBA 7.6.6) template. It is
731 a list of VLs corresponding to SLs 0-15 (Note
734 Typical default values (hard-coded in OpenSM initialization) are:
751 qos_ca_ - QoS configuration parameters set for CAs.
752 qos_rtr_ - parameters set for routers.
753 qos_sw0_ - parameters set for switches' port 0.
754 qos_swe_ - parameters set for switches' external ports.
764 off-subnet DGIDs. By default, the SA fails such queries.
765 Note that IBA does not specify how the SA should obtain off-subnet path
767 The prefix routes configuration is meant as a stop-gap until the
770 Each line in the configuration file is a 64-bit prefix followed by a
771 64-bit GUID, separated by white space.
777 Either, or both, can be wild-carded by specifying an
780 When responding to a path record query for an off-subnet DGID,
783 a wild-carded prefix at the beginning of the configuration file renders
789 are wild-carded means that a path record query specifying any
790 off-subnet DGID should return a path to the first available router.
792 compiling opensm with -DROUTER_EXP which has been obsoleted.
801 m_key - the 64-bit MKey to be used on the subnet
803 m_key_protection_level - the numeric value of the MKey ProtectBits
805 m_key_lease_period - the number of seconds a CA will wait for a
811 Switches and HCAs with a non-zero MKey will not accept requests to change
824 0 - Queries return valid data, including MKey. Configuration changes
826 1 - Like level 0, but the MKey is set to 0 (0x00000000) in queries,
828 2 - Neither queries nor configuration changes are allowed, unless the
830 3 - Identical to 2. Maintained for backwards compatibility.
839 (Bad M_Key, Trap 256). If the MKey lease period is non-zero, it also starts a
851 any running timers. For maximum protection against accidentally-exposed MKeys,
856 re-enabled with an interval less than the Mkey lease period.
875 1. Min Hop Algorithm - based on the minimum hops to each node where the
878 2. UPDN Unicast routing algorithm - also based on the minimum hops to each
883 3. DNUP Unicast routing algorithm - similar to UPDN but allows routing in
887 4. Fat Tree Unicast routing algorithm - this algorithm optimizes routing
888 for congestion-free "shift" communication pattern.
890 fat-tree of various types, not just K-ary-N-Trees: non-constant K, not
894 5. LASH unicast routing algorithm - uses Infiniband virtual layers
895 (SL) to provide deadlock-free shortest-path routing while also
897 deadlock-free topology-agnostic routing algorithm to the non-minimal
900 6. DOR Unicast routing algorithm - based on the Min Hop algorithm, but
906 7. Torus-2QoS unicast routing algorithm - a DOR-based routing algorithm
907 specialized for 2D/3D torus topologies. Torus-2QoS provides deadlock-free
913 8. DFSSSP unicast routing algorithm - a deadlock-free
914 single-source-shortest-path routing, which uses the SSSP algorithm
916 Infiniband virtual lanes (SL) to provide deadlock-freedom.
918 9. SSSP unicast routing algorithm - a single-source-shortest-path routing
950 c. if none - prefer those which go through another NodeGuid
956 the fabric switches unless the -r (--reassign_lids) option is specified.
960 --reassign_lids
962 end nodes. Specifying -r on a running subnet
964 Without -r, OpenSM attempts to preserve existing
975 to recheck LIDs (by GUID) for disconnected nodes, and LFTs for non-existent
983 specified. It can also be invoked by specifying '-R minhop'.
986 min-hop tables on every switch and LFT output port assignment. Link
990 -i <equalize-ignore-guids-file>
992 \-\-ignore_guids <equalize-ignore-guids-file>
1005 of the subnet. A loop-deadlock is a situation in which it is no longer
1013 1. Auto-detect root nodes - based on the CA hop length from any switch in
1018 the root nodes found by this auto-detect stage is used by the ranking
1026 2. Ranking process - All root switch nodes (found in stage 1) are assigned
1029 rules that ensure loop-free paths.
1031 3. Min Hop Table setting - after ranking is done, a BFS algorithm is run from
1036 At the end of the process, the updated FDB tables ensure loop-free paths
1051 Use '-R updn' option (instead of old '-u') to activate the UPDN algorithm.
1052 Use '-a <root_guid_file>' for adding an UPDN guid file that contains the
1054 If the `-a' option is not used, OpenSM uses its auto-detect root nodes
1060 format will be discarded.
1079 Fat-tree Routing Algorithm
1081 The fat-tree algorithm optimizes routing for "shift" communication pattern.
1083 fat-tree of various types.
1084 It supports not just K-ary-N-Trees, by handling for non-constant K,
1086 As in UPDN, fat-tree also prevents credit-loop-deadlocks.
1088 If the root guid file is not provided ('-a' or '--root_guid_file' options),
1089 the topology has to be pure fat-tree that complies with the following rules:
1090 - Tree rank should be between two and eight (inclusively)
1091 - Switches of the same rank should have the same number
1092 of UP-going port groups*, unless they are root switches,
1093 in which case the shouldn't have UP-going ports at all.
1094 - Switches of the same rank should have the same number
1095 of DOWN-going port groups, unless they are leaf switches.
1096 - Switches of the same rank should have the same number
1097 of ports in each UP-going port group.
1098 - Switches of the same rank should have the same number
1099 of ports in each DOWN-going port group.
1100 - All the CAs have to be at the same tree level (rank).
1103 fat-tree, and it should only comply with the following rules:
1104 - Tree rank should be between two and eight (inclusively)
1105 - All the Compute Nodes** have to be at the same tree level (rank).
1106 Note that non-compute node CAs are allowed here to be at different
1112 ** list of compute nodes (CNs) can be specified by \'-u\' or \'--cn_guid_file\'
1117 to no longer be "pure" fat-tree.
1119 Note that although fat-tree algorithm supports trees with non-integer CBB
1125 pure and symmetrical fat-tree, the more optimal the routing will be.
1127 The algorithm also dumps compute node ordering file (opensm-ftree-ca-order.dump)
1132 Routing between non-CN nodes
1134 The use of the cn_guid_file option allows non-CN nodes to be located on different levels in the fat…
1135 In such case, it is not guaranteed that the Fat Tree algorithm will route between two non-CN nodes.
1136 To solve this problem, a list of non-CN nodes can be specified by \'-G\' or \'--io_guid_file\' opti…
1137 …itches the wrong way round a specific number of times (specified by \'-H\' or \'--max_reverse_hops…
1140 Please note that using max_reverse_hops creates routes that use the switch in a counter-stream way.
1149 Use '-R ftree' option to activate the fat-tree algorithm.
1150 Use '-a <root_guid_file>' to provide root nodes for ranking. If the `-a' option
1152 Use '-u <root_cn_file>' to provide the list of compute nodes. If the `-u' option
1155 Note: LMC > 0 is not supported by fat-tree routing. If this is
1163 agnostic deadlock-free routing within communication networks.
1166 topology for the shortest-path routes between all pairs of sources /
1176 1) LASH determines the shortest-path between all pairs of source /
1178 SRC/DST - DST/SRC pairs and there is no guarantee that the return
1204 related to a root node and always routes shortest-path.
1209 Use '-R lash -Q ' option to activate the LASH algorithm.
1224 invoke this use '-R lash -Q --do_mesh_analysis'. This will
1237 dimension or a mesh dimension. Alternatively, the -O option can be
1246 hypercube dimension and match on both ends of the cable, or the -O
1250 along the mesh dimension, or the -O option used as an override.
1252 Use '-R dor' option to activate the DOR algorithm.
1256 The (Deadlock-Free) Single-Source-Shortest-Path routing algorithm is
1259 uses Infiniband virtual lanes (SL) to provide deadlock-freedom.
1265 network and each edge represents one direction of the full-duplex
1302 deadlock-free and an relation table is generated, which contains
1307 This algorithm does not perform the steps 3)-5) and can not be
1308 considered to be deadlock-free for all topologies. But on the one
1310 (5,000+ CAs and deadlock-free by design) to reduce
1312 the SSSP routing algorithm as an alternative, when all deadlock-free
1320 a) running DFSSSP: '-R dfsssp -Q'
1328 b) running SSSP: '-R sssp'
1337 configuration CN1-CN3 might send all I/O traffic via Link2 to IO1,IO2:
1341 \\ /----\\ /
1343 CN2 -- Switch1 Switch2 -- CN4
1345 / \\----/ \\
1350 file and the I/O guid file specified by the \'-u\' or \'--cn_guid_file\' and
1351 \'-G\' or \'--io_guid_file\' options (similar to the Fat-Tree routing).
1358 compute nodes -> I/O nodes -> other nodes
1362 a) neither \'-u\' nor \'-G\' are specified: all nodes a treated as \'other nodes\'
1365 b) \'-G\' is specified: traffic towards I/O nodes will be balanced optimally;
1368 but the balancing focus should be I/O, then one has to use \'-u\' and \'-G\'
1374 Torus-2QoS Routing Algorithm
1376 Torus-2QoS is routing algorithm designed for large-scale 2D/3D torus fabrics;
1377 see torus-2QoS(8) for full documentation.
1379 Use '-R torus-2QoS -Q' or '-R torus-2QoS,no_fallback -Q'
1380 to activate the torus-2QoS algorithm.
1385 To learn more about deadlock-free routing, see the article
1402 "Routing for the ASI Fabric Manager", Solheim et al. IEEE
1412 J. Domke, T. Hoefler and W. Nagel: Deadlock-Free Oblivious Routing
1417 Large-Scale InfiniBand Networks, In 17th Annual IEEE Symposium on High
1428 One existing routing module is up-down "updn", which may be
1429 activated with '-R updn' option (instead of old '-u').
1432 $ opensm -R 'module-name'
1439 - this will load switch LFTs and/or LID matrices (min hops tables)
1440 - this will load switch LFTs according to the path entries introduced
1442 - no additional checks will be performed (such as "is port connected",
1444 - in case when fabric LIDs were changed this will try to reconstruct
1449 The file format is compatible with output of 'ibroute' util and for
1454 opensm -R file -U /path/to/lfts_file
1464 name should be specified by -M or --lid_matrix_file option. For example:
1466 opensm -R file -M ./opensm-lid-matrix.dump
1468 The dump file is named \'opensm-lid-matrix.dump\' and will be generated
1478 Both or one of options -U and -M can be specified together with \'-R file\'.
1487 The per module logging config file format is a set of lines with module
1498 ---- -----------------
1499 0x01 - ERROR (error messages)
1500 0x02 - INFO (basic messages, low volume)
1501 0x04 - VERBOSE (interesting stuff, moderate volume)
1502 0x08 - DEBUG (diagnostic, high volume)
1503 0x10 - FUNCS (function entry/exit, very high volume)
1504 0x20 - FRAMES (dumps all SMP and GMP frames)
1505 0x40 - ROUTING (dump FDB routing information)
1506 0x80 - SYS (syslog at LOG_INFO level in addition to OpenSM logging)
1514 .B /etc/opensm/ib-node-name-map
1515 default node name map file. See ibnetdiscover for more information on format.
1522 .B /etc/opensm/qos-policy.conf
1526 .B /etc/opensm/prefix-routes.conf
1530 .B /etc/opensm/per-module-logging.conf
1534 .B /etc/opensm/torus-2QoS.conf
1535 default torus-2QoS config file
1561 torus-2QoS(8), torus-2QoS.conf(5).