Data loss

These days we are hearing more often about data loss events at the WLCG sites. Today it was the NL-T1 site that reported some data loss in the daily Operations meeting. Apparently, the tape drive loaded the tape and, instead of reading it, it just destroyed it. A similar event happened to us at PIC at the end of September, when we lost a tape containing 214 files from CMS. Nothing could be done with that piece of hardware... not even rewinding it! Luckily for us, all of those files were replicated in some other Tier-1 or at CERN, so we could fix the problem quite straightaway.
We were used to think in tapes as a safe media for data... but these episodes of tape destruction show that this is not always the case. A bit scary.
Anyhow, even ig it does not solve anything but it is nice to see that WLCG sites we are not the only ones losing data. Even Microsoft loses some data eventually!

August availability: on target, but...

Last August, as may be one could expect after a busy July with the CPU occupied at almost 100%, many of the LHC experiment production managers and computing guys went on (deserved) holidays. Accounting data is being collected these days, and the results for PIC show that only about 40% of the installed CPU was used. Curiously, the experiment share of this workload was highly non-nominal: LHCb, which represents only about 10% of PIC pledges, was the one consuming more: up to 60% of the delivered CPU cycles were for them. ATLAS got essentially all the remaining 30%, while CMS did essentially zero. So, for both ATLAS and CMS August was the month with the lowest CPU consumed in the year, while for LHCb was a record-breaking month.
PIC Tier-1 availability during August was just right on top of the target: 97%. About 1% of the unavailability was due to the usual monthly Scheduled Downtime which took place on the 25th August. Most of the remaining 2% unavailability we spent it also on that same day, which suggests that there is still room for improvement on SD coordination. One of the A-critical services, the site-bdii, was off during almost 4h after its scheduled intervention... and no one of us noticed! There was also an issue with the Computing Service (B-critical) which had its queues closed for 2h longer than planned. We should now then feed this experience back into our operation system and make sure the relevant procedures are improved.
Besides that, the 19th of August instabilities appeared in the OPN link which were affecting the SRM service, specially for outgoing transfers. The problem disappeared in about 24h, but we never knew what had really happened. The Spanish NREN did not answer to our query for information. First we thought this was an August-effect, but later we realised the problem was that our e-mail contact for operational issues in the network was wrong. We have corrected this and the e-mail we have now should even trigger a ticket opening automatically.
The good news for August were that, despite being one of the hottest in several years, the cooling system of the PIC machine room coped perfectly with it. Seems that the new maintenance team did a good job in preparing the system for the summer campaign.

One Petabyte knocking at the door

This is just a micro-post, to show you who I found two days ago knocking at PIC's door just when I was leaving the office.
Yes! the long-awaited Petabyte that will push our capacity up to the 2009 MoU pledges was there.
It will be a tricky path the one we have to walk from these nice wooden boxes to the WLCG SRM service (the first one seems to be the big disks are reluctant to get to Barcelona).
Busy weeks ahead... but the goal is clear: fill up those disks with LHC data.

Believe us, PIC is Ok

Scary, isn't it? PIC availability has been bully red in the last 24h but, sadly, there was not much we could do.
We have always said, and still strongly believe it, that SAM tests are a very good thing. Actually, I firmly believe that have been one of the key ingredients for the WLCG success. Success here meaning the evolution from "the Grid does not work" situation with 60% job success rate we had few years ago, to the rutine >97% availabilities we are used to see these days. But yes, not even SAM tests are perfect. There has always been a dark corner inside them: the so much questioned "SE test inside the CE", or lcg-rm test. And inside this controversial test there is another smaller corner which is still a bit darker: the file replication to CERN test. This was the one that started flickering on Tuesday at PIC and it is consistently failing since more than 24h. This test tries to copy a file sitting at PIC into a very concrete DPM server at CERN. This very precise connection was timing out for us while any other transfer to any other site, even to any other CERN storage server was working. This was strange enough so that we asked for help to our CERN colleagues. Today, they came with the good news: problem found, a problematic router.
Got a really puzzling error? Bet on the network...

July: good availability comes with CPU delivery record

Most of the people is these days either just back from holidays (me) or still out (lucky them) or neither of those (also lucky, since they will leave later...). Anyhow, life at the Tier-1 is 24x7 since we all know so, it doesn't matter if we are in the middle of August and it is near 40 degrees celsius out there, it is time to report about service performance in the past month.
We just got the WLCG reports for July and the results for PIC are pretty positive: 99% availability and reliability. This little 1% that gets us away of our beloved 100% happened on the 29th July, around noon. During four hours all of the Computing Elements at PIC were failing the SAM Job Submission tests, so definetely the Tier-1 service was affected during that period. The source of the problem was found to be a pretty mysterious one: the switch connecting the servers hosting Virtual Machines to the PIC LAN (VMs are always a bit misterious, aren't they?). Actually, the source of the problem was not found, but just disappeared when that switch was replaced by a new one (different brand, no names here to avoid anti-propaganda :)
Regarding the availability of PIC as seen from the experiment specific monitoring, we got also very good results for ATLAS and LHCb, close to 100%. However, the result for CMS was not that good: a mere 90%. This funny assymetry was due to a bug we introduced by mistake in the Torque ACL queues configuration which actually blocked CMS submissions to our short queue for 3 days (6-9 July). Somebody could ask why we did not notice this in 3 days... We should put priority in deploying the WLCG-Nagios in production. It will for sure help reducing these unavailable times.
So, besides this unfortunate CMS-blocking bug, we can say July was a pretty good month for the Tier-1 in terms of availability. Thanks to this, and also to the job submission hyperactivity seen from ATLAS and LHCb, we delivered a record amount of CPU cycles during July: around 70.000 ksi2k·days, which is very close to keeping 100% of our resources busy for the whole month.
Now things look quiet, being many people still away. PIC is up and running, cooling is ok (even the heat wave out there)... but watch out for the VM-networking ghosts, they could come at any time.

Welcome Valparaiso !

The Universidad Técnica Federico Santa Maria (UTFSM) of Valparaiso (Chile) has joined ATLAS Computing and has been associated to PIC as its Tier-1 center. They are now a small center (Tier-3) but surely will grow soon. Firsts transfers using the full chain of the ATLAS Data Management System has been successfully tested. We welcome UTFSM to PIC and to the distributed computing world !

WLCG proposed a test period were all the VOs can exercise their computing models altogheter in a realistic scenario before the data taking. This exercise was named STEP09. The challenge ended on Friday 23:00 CET. A lot of activity happened during these last two weeks and the ATLAS Computing Model were fully exercised (finally!):

a) Huge load of MC production, exercising the farms and the data aggregation form T2s to the T1. The job type were basically G4 simulation jobs and AOD merging.

b) Huge load of Analysis jobs that were racing against the MC ones. I'd like to emphasize that this activity was quite well *disorganized* in overall... that was good as possibly could simulate quite well the possible impact of the uncontrolled user analysis that everyone will get very soon. There were general problems with Athena build jobs (compatibly libraries), the PanDA brokerage were bypassing the ATLAS release check at the sites (spotted last Thursday), the protocols to get data were basically lcg-cp for File stager and dcopen/file for remote connection... there's a big room for improvement in the Pilot Jobs based User Analysis, hopefully some tests will be performed in the coming months (use "natural SE" protocols for the FS, tune the Read Ahead Buffer -32Kb used during STEP-, etc.)

c) Non-Stop Reprocessing, this was the primary activity at the Tier-1s and was focused to perform multi-VO reprocessing at the same time fro those Tier-1 serving more than one experiment. That was achieved: ATLAS, CMS and LHCb managed to use the robots at the same time during several days. Six ATLAS sites manage to pass the metrics: reprocess at five times the data taking speed assuming 40% of LHC machine efficiency. And three sites manage to get gold stars, the metrics for this were: reprocess at five times data taking speed assuming 100% of LHC machine efficiency. PIC obtained gold star.

e) Data distribution: was broadly tested on top of the processing activity. Merged AOD were pre-placed at the sites: T1-T1-T2s, Functional Tests kept running all the time during the two weeks. There were no major incidences although some sites had some instabilities during the STEP: disk space emptied out, gridftp doors overloads (intensive LAN, WAN activity), but in overall data flow has shown its robustness and major improving with respect two years ago.

The STEP showed that manual work is still required but we managed to spot and learn which things would be the most demanding ones, the target now is to work in this direction and deploy an scalable operations model before the start.

Concerning our cloud, I'd say the exercise was a real success ! I want to thanks every party involved. Site has been stable (no major issues) that allowed STEP to pass through in a quite intensive way. I've barley seen an idle CPU all over the cloud, letting schedulers to battle against the different roles and jobs. Also the data transfer has been pretty high and the output form PIC to the Tier-2s has been quite stable at around 150-200MB/s.




MonteCarlo production ran between 1k and 1.5k jobs during STEP and efficiencies reached were well higher the 90%




On the other side User Analysis jobs were also filling the remaining slots, see the snapshot for the Pilot-based jobs (the ones directly submitted trough WMS are not accounted). Efficiencies ranked among the 70% and 80% depending on the sites playing the game, there were some issues (all understood!) that prevented pilot user analysis jobs to finish correctly.






PIC reprocessed the data two times, job efficiencies and data flow from the robot to the buffers and to the WN as a final step showed a very good performance. Pre-Stage is manage to keep up with 500 concurrent jobs which is beyond the required reprocessing speed for a Tier1 of small size. The interesting point is that reprocessing outputs were written to tape as well, this is what the computing model say, and we found no major problems with the simultaneous usage of drive for reading and writing.





What we learned during the STEP is that, once again, storage services are the most sensitive layer of the infrastructure. Small outage on the SE induce a general failure of every single activity. Storage services would have to be well dimensioned on those sites that had instabilities under constant load. We learned also that disk space is as volatile as it is in our laptops, a good prevision of the storage versus desired data is mandatory, it is true that ATLAS has to be more aggressive in data deletion but this is at the hands of the physics groups. I'd say that we should reinforce the Federative structure of our Tier-2s and take profit of these for the data distribution and always preserve some disk quantity in case of crisis. In the near future ATLAS will provide an interface that for modifying the shares at the sites so can be dynamic and not human-based, but also DDM will cross-check if the space token has enough free space before the data is shipped, hence preventing to replicate in case of shortages.

Once again my feeling and understanding is that our cloud is ready for data taking, let's hope data start to flow soon...