[Halld-online] [New Logentry] Follow-up Re: Data files not being copied to tape library

[davidl at jlab.org]

Logentry Text:
--
It turns out the data files from gluonraid2 actually are now being copied. The problem is likely due to our data rate being larger
than the capability of the system for copying to tape. It took a few hours for the gluonraid2 files to rise to the top of the job queue.

Looking at creation times of 17 files from run 2145 on the tape library (via creation times of the stub files) and the creation times of the same 17 files on the RAID disk,
I can see they span 24 minutes on the former and only 19 minutes on the latter. This means we are writing to disk about 25% faster than we can copy to tape. This
is consistent with the numbers from Paul Letta's tests this summer indicating we could copy to tape at ~450MB/s and what we know about our current data rate (in mode 8)
being ~600MB/s.

Given this difference, we are filling up the disk faster than we can empty it at a rate of about 150MB/s. This means we will be about 1 RAID disk behind in about 5.5 days.

We will probably need to consider changing the ratio of mode 8 to mode 7 data taking. From data taken earlier today, the average event size for mode 7 events is a factor of 3 to 4 smaller than mode 8 data. If we spend 50% of wall time in mode 7 rather than mode 8, then the copy to tape should just barely be able to keep up.

---

This is a plain text email for clients that cannot display HTML.  The full logentry can be found online at https://logbooks.jlab.org/entry/3313596
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mailman.jlab.org/pipermail/halld-online/attachments/20141213/2181c574/attachment.html