This document describes the configuration of the SAM products via the server list file.
The server list file configures the various SAM products and can be found in general in "~sam/private". It is usually called "hostname_server_list.txt"
Please refer also to my description of how to configure SamCp which has some overlap with this page
N.B. There is one line per product which is configured (although this might get
The general layout of a configuration line is
product config version station-name parameters
where:
| product | config | version | station-name | parameters |
|---|---|---|---|---|
| product to configure, e.g. station | configuration taken vom sam_config, usually prd | version to set up, eg. v4_2_1_63 | name of the sam station, e.g. cdf-fzkka | parameters described below |
The description below is based on my present knowledge and hence incomplete, however it is a working example from GridKa, Karlruhe, Germany
This section will describe the parameters used for configuring the station itself
| parameter | description | example | comment |
|---|---|---|---|
| -OAhost | tells the Orbacus part which host the station runs on | -OAhost cdf.fzk.de | not needed in general |
| -OAport | tells the Orbarcus part which port to use | -OAport 6789 | not needed in general |
| --min-delivery | minimal file-size which has to be exceeded before file transfer is started | --min-delivery=1k | |
| --pmaster-arg=--consumption-map= | tells the roject master which workernodes are accepted as "own" worker nodes and are allowed to run projects with this station | --pmaster-arg=--consumption-map=\.\*::cdf.fzk.de | Note the regular expression: "\.\*::cdf.fzk.de" The first part determines the domain (i.e. output of hostname -f) of the worker nodes, the second part the fully qualified hostname of the node running the station. In this example, the station running on cdf.fzk.de accepts all worker nodes. Note that this is an argument of the project master and hence has to be passed on by the station to each new project master. |
| --log-file | determines prefix to a per-day log file in addition to the trace file | --log-file=station.log | creates a file called "station.log__date" once per day which has the same content as the trace file. Unlike the tracefile, this logfile is not overwritten upon station restart |
| --max-prefetched-files | determines the number of files the station is allowed to prefetch, i.e. when the job on the worker nodes are consuming all presently cached files, the station will retrieve n further files | --max-prefetched-files=10 | please do not exceed this as we have encountered problems with the GSI enabled doors at FNAL this year |
| --preferred-loc | determines where to look first for files | --preferred-loc=/pnfs/cdfen/filesets | first go to Enstore at FNAL This option accepts a comma-separated list |
| --omit-loc | determines where not to look for files | --omit-loc=dcap | This option accepts a comma-separated list |
| --common-timeout | general timeout in minutes | --common-timeout=180 | Timeouts have to be determined by monitoring typical situations for some time, please do not just copy this value but monitor your station to determine a value suitable for you |
| --intrastation-timeout | timeout for intra-station transfers (?) | --intrastation-timeout=180 | Timeouts have to be determined by monitoring typical situations for some time, please do not just copy this value but monitor your station to determine a value suitable for you |
| --routing-station | if a file cannot be accessed directly (e.g. stken), another station has to retrieve the file on behalf of the requesting station first | --routing-station=enstore::cdf-sam | Please refer to how to configure SamCp |
| --routing-user | determines the username which will appear on the routing station | --routing-user=cdf-fzkka | Please use the name of the SAM station here to make identification easier Please refer to how to configure SamCp |
| --routing-group | determines the group used for the routing request | --routing-group=test | Always use test here Please refer to how to configure SamCp |
The File Storage Manager (FSS) is respoinsible for storing files, i.e. accept a file from the user when a "sam store" command is issued, transfer the file to the station at which the file is going to be stored, etc.
| parameter | description | example | comment |
|---|---|---|---|
| -OAhost | tells the Orbacus part which host the station runs on | -OAhost cdf.fzk.de | generally not needed |
| --route | determines the route, i.e. if a storage area cannot be accessed directly, which station to contact to store the file | -route=enstore,cdf-sam:fcdfdata016.fnal.gov:/data1/permanent/prd |
The stager handles the disks attached to a station
| parameter | description | example | comment |
|---|---|---|---|
| --with-fss | register with FileStorageManager | ||
| --without-fss | do not register with FileStorageMaganer | ||
| --with-sm | register with station | if the station does not spawn stagers on its own for the caches, they have to be started manually with this option. Note You need one stager per cache and tell the station not to spawn own stagers (but I have forgotten how to do that...) |
|
| --without-sm | do not attach to station | The station spawns own stagers to manage its caches | |
| --max-transfers | max. number of simultanious file transfers | --max-transfers=5 | This number is per stager i.e. if several stagers run, the total number of simultanious file-transfers is n*number_of_stagers |
| -OAhost | tell the Orbacus part which host the station runs on | -OAhost cdf.fzk.de | not needed in general |