aLIGO LHO Logbook

H1 CDS

david.barker@LIGO.ORG - posted 14:43, Friday 21 June 2013 - last comment - 20:02, Friday 21 June 2013(6844)

started first EX front end computer, caused DAQ problems, stopped running computer

Following the installation of timing, networking and ipc in EX, we booted the h1susauxex front end computer and started the IOP model. This caused mx-stream DAQ errors from many other front ends, including all at EY and on the LVEA test stand. Restarting the mx streamers on these front ends did not clear the problem. When h1susauxex was powered down (thereby killing the mx streaming) all the other front ends became good again with intervention needed.

We will keep EX powered down for this weekend and revisit this problem Monday or Tuesday next week.

Comments related to this report

keith.thorne@LIGO.ORG - 14:54, Friday 21 June 2013 (6845)

Link

You may be encountering problems when manually starting models
 See this wiki page https://awiki.ligo-wa.caltech.edu/aLIGO/FrontEndStart

You will see that the old scripts start_streamers.sh and start_iop_mx_stream.sh still use a single port MX stream. 
As described, modifying start_streamers.sh is easy (merely stop/start /etc/init.d/mx_stream). Discard start_iop_mx_stream.sh.

* I have tested the new startup scheme described there on the LLO DAQ test stand, but have not attempted in on L1.
The RCG 2.7 roll-out would be a good chance to do that.

keith.thorne@LIGO.ORG - 19:32, Friday 21 June 2013 (6853)

Link

Also make sure that the /diskless/root/etc/rtsystab file has been updated with a line for the new computer and IOP model.  Several startup scripts depend on it.  In particular /etc/init.d/mx_stream uses it to figure out the MX stream card and slot.

cyrus.reed@LIGO.ORG - 20:02, Friday 21 June 2013 (6854)

Link

There was a configuration error on the EX DAQ switch on two ports, including the one for h1susauxex (h1seiex being the other).  It was missing the mtu setting to allow transmission of jumbo frames (the dainbramaged CLI pagination code in these switches makes this hard to see, even when expressly looking for it...).  This has been corrected, and I suspect it will work now barring any other problems, but this is yet to be tested.