Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.atnf.csiro.au/vlbi/dokuwiki/lib/exe/fetch.php/difx/meetings/hobart2015/rottmann_mark6tools.pdf Дата изменения: Wed Nov 18 05:36:13 2015 Дата индексирования: Tue Apr 12 12:44:32 2016 Кодировка: Поисковые слова: cygnus

DiFX-Mark6 integration
correlator operations & tools

Helge Rottmann
Max-Planck-Institut fЭr Radioastronomie
9th DiFX Users and Developer Meeting Hobart, Tasmania


Mark6 Overview



Mark6 has 4 banks Modules contain up to 8 disks Module grouping When recording a group can be formed of up to 4 modules Data is scattered over all disks of the group Complete module group has to be present on playback


Mark6 disk devices


The individual disks contain a normal FS (default XFS) The individual disks are normal linux disk devices =>Disks must be mounted prior to reading the data




Standard Mark6 FS structure
Mark6 disk device contains two partitions:
1st partitition contains vdif scan data
/data /c22gl_Jc_081-0500.vdif /c22gl_Jc_081-0528.vdif ...

2nd partition contains meta data
/eMSN /disk_sn /group /slist


Mark6 metadata
/eMSN
BHC%0033/48008/4/8

/serial_sn
0:AR11021EHBWVUB 1:AR11021EHEAXWB 2:AR11021EHEDRDB 3:AR11021EHEBGGB 4:AR11021EHE2DPB 5:AR11021EHE1GZB 6:AR11021EHEBJPB 7:AR11021EHEDR0B

/group 4:BHC%0033/48008/4/8:BHC%0034/48008/4/8:BHC%0035/48008/4/8: BHC%0036/48008/4/8 /slist
{1: {'status': 'recorded', 'num_str': 'create_time':'2015y079d05h29m22s', 's 'spc': 0, 'size': '1.028'}, 2: {'status': 'recorded', 'num_str': 'create_time':'2015y079d16h50m24s', 's 'dur': 1, 'spc': 0, 'size': '1.028'}, 1, 'start_tm': 1426829460, n': 'bbb_ccc_aaa', 'dur': 1, 2, 'start_tm': 1426870320, n': 'fringe_test_local', ...


Current DiFX-Mark6 operations
DiFX-Mark6 operations at the moment:


Look up module group members (e.g. from logs) Load and manually mount the module disks Create .vsum content summary Modify .v2d, e.g.
machine= mark6-01 mark6filelist = station.vsum









Unmount module disks when done => extending mark5daemon / genmachines to automatically do all this


Mark6 default mount structure
mount structure:
/mnt/disks/[1234]/[01234567]/data /mnt/disks/.meta/[1234]/[01234567]

=> bank/disk position of a device needs to be correctly determined


Mark6 mount complications
1) Device order of a single module is ,,random" when key is turned
2) Devices from multiple modules can become ,,mixed" 3) (Theoretically) non-Mark6 hotplug devices can appear 4) Discovery & mounting of new devices can take some time. Must check module completness before starting correlation Solutions: 1), 2) and 3) can be resolved by making use of udev 4) can be resolved by inspecting metadata


mk5daemon
Requirements:


Identifiy bank/disk position of each device upon key turn Mount/unmount module disk devices upon key turn Parse meta data to automatically determine e.g. module group, scan list, expected number of disks etc. Perform sanity checks (e.g. module completness, group completness etc.) Send out difxmessage (status)










mk5daemon: current state
mk5daemon coding (nearly) done and tested > mk5daemon -6
code exists in branch /applications/mk5daemon/branches/mk6




To do items:
Merge with trunk (hopefully during this meeting) Move code parts to a new mark6 library Automatically create .dir/.vsum/filelist files from metadata?






genmachines
Requirements:


cluster definition file to contain mk6:// tag parse Mark6StatusMessage
­ ­

Validate module group completness Assign Mark6 in .machine file


mk6mon


similar to mk5mon display bank content of a mark6 unit and show current state, playback bandwidth etc. implemented



To do items:


Indicate module group /disk completness


Mark6StatusMessage


sks> sks> sks> sks>


Other considerations
Vex
use $TAPELOG_OBS section to specify modules?