FNAL: Had some downtime due xrootd being off on one data server. Caused clients to get dreaded 84 error.
Monitoring had downtime due to expired proxy.
Frontend at UCSD is down. Waiting for new hardware - can turn it back on in the VM, but might have to shut it off again at any given time. Likely in the next 2 weeks (possibly in the next few days).
Production redirector had hostcert expire.
Gratia went down due to unrelated OOM issues.
Matevz's collector has held up.
Florida had issues with Lustre, but Xrootd responded correctly.
Software:
Dan has been working on the read-ahead patch for the XrdHdfs plugin. Being tested in Wisconsin and lightly at Nebraska.
IOPS to HDFS halves.
Re-use of bytes in the buffer is about 10%.
Brian: What percentage of "useless" bytes?
Dan: Been looking at the CVMFS design. Might be useful regardless locally for $OSG_APP. Not looking at the code yet.
Igor: How would you handle "committing" file modifications to the file tree to make them available? Dan: Might be a lot of concurrency/race issues here to have a "cronjob" do the updates for us.
Brian: Might want to think about doing a FUSE-emulation layer for Parrot.
Brian: Summary of Xrootd meeting (will put a pointer to slides)
AOB:
Ken: Can we do an update at the T2 meeting?
Brian: Would be nice to have Matevz do it, but don't want to volunteer him when he's on vacation. Dan will be backup if Matevz is unable.
Topics: Job Overflow and monitoring
Igor: Student we were hoping to use to look at data placement has disappeared; will have to find a new one. UCSD has candidate for job position; might free up Igor/Jeff/Matevz's time.
Brian: At the end-stages of hiring new personnel.
Dan: Got a new student working on Parrot and CMSSW.