Generates a topology XML file of a cluster using topology.csv, linksum_swd06.csv, and linksum_swd24.csv as input. The topology file can be used to bring up and verify the cluster.
Produces full help text.
Specifies the output detail level. Default = 0. Levels are additive.
By default, the top level is always produced. Switch, rack, and rack group topology files can be added to the output by choosing the appropriate level. If the output at the group or rack level is specified, then group or rack names must be provided in the spreadsheet. Detailed output can be specified in any combination. A directory for each topology XML file is created hierarchically, with group directories (if specified) at the highest level, followed by rack and switch directories (if specified).
Intel® Omni-Path Edge Switch 100 Series topology files.
Rack topology files.
Rack group topology files.
Specifies the verbose level. Range = 0 - 8. Default = 2.
No output.
Progress output.
Reserved.
Time stamps.
Reserved.
Specifies the output indent level. Range = 0 - 15. Default = 0.
Specifies DO NOT clean temporary files.
Prevents temporary files in each topology directory from being removed. Temporary files contain CSV formatted lists of links, HFIs, and switches used to create a topology XML file. Temporary files are not typically needed after a topology file is created, however they are used for creating linksum_swd06.csv and linksum_swd24.csv files, or can be retained for subsequent inspection or processing.
Used on Multi-Rail or Multi-Plane fabrics. Can be used to override the default hfi1_0.
For Multi-Plane fabric, use the tool multiple times with a different hfi-suffix.
For Multi-Rail fabric, specify HostName as "HostName HfiName" in the spreadsheet.
The opaxlattopology script reads the topology.csv file from the local directory, and reads the other files from /usr/share/opa/samples/linksum_swd06.csv and /usr/share/opa/samples/linksum_swd24.csv. The topology.csv file is created from the topology.xlsx spreadsheet by saving the Fabric Links tab as a .CSV file to topology.csv. A sample topology.xlsx is located in the /usr/share/opa/samples/ directory. Inspect the topology.csv file to ensure that each row contains the correct and same number of comma separators. Any extraneous entries in the spreadsheet can cause the CSV output to have extra fields.
The script outputs one or more topology files starting with topology.0:0.xml. The topology.csv input file must be present in the same directory from which the script operates, but the linksum CSV files are read from the /usr/share/opa/samples directory.
opaxlattopology # reads default input 'topology.csv' and creates default # output 'topology.0:0.xml' opaxlattopology fabric_2.csv # reads input 'fabric_2.csv' and creates default output
See topology.xlsx for examples of links between HFI and Edge SW (rows 4-7), HFI and Core SW (rows 8-11), and Edge SW and Core SW (rows 12-15).
The following environment variables allow user-specified MTU.
If set, it overrides default MTU on switch-to-switch links. Default = 10240
If set, it overrides default MTU on switch-to-HFI links. Default = 10240
The linksum_swd06.csv and linksum_swd24.csv files are provided as stand-alone files in the /usr/share/opa/samples directory. However, they can be recreated (or modified) from the spreadsheet, if needed, by performing the following steps:
By default, the opaxlattopology script considers all of the hosts to have a single HFI (hfi1_0).
For Multi Rail/Plane fabrics, the user has following options:
For example, if there are two identical fabric (fabric_1 and fabric_2) connected to a single host with two HFIs (hfi1_0 and hfi1_1), the tool can be run twice like this:
For fabric_1:
opaxlattopology topology.csv topology.xml
For fabric_2:
opaxlattopology -s hfi1_1 topology.csv topology.xml