TChain Buffer issues - disk related?

ROOT
1

Hi

Unfortunately I cannot reproduce my problem with a short simple example so I appreciate this is not the best question but I would really appreciate any advice on the problem.

Basically when I TChain ~ 5000 files with ~3M events and then loop over all the tracks in the events and do some event mixing jobs sometimes (not always) fail with warning and errors from the TBasket, TBuffer and then result in a seg-fault. Typical output looks like:

[quote]Warning in TClass::TClass: no dictionary for class StFlowPicoEvent is available
Warning in TClass::TClass: no dictionary for class StFlowPicoTrack is available
(incomplete l-tree) R__unzip: error during decompression
Error in TBasket::ReadBasketBuffers: fNbytes = 9019, fKeylen = 83, fObjlen = 15196, noutot = 0, nout=0, nin=8424, nbuf=15196
Warning in TBasket::ReadBasketBuffers: basket:fTracks.mPidElectron has fNevBuf=9, but fEntryOffset=0, trying to repair
Error in TBranchElement::GetBasket: File: /auto/rnc3/sarblyth/production62/Days/088//st_physics_5088046_raw_1030020.flowpicoevent
.root at byte:4556260, branch:(null), entry:150153100
(incomplete l-tree) R__unzip: error during decompression
Error in TBasket::ReadBasketBuffers: fNbytes = 9019, fKeylen = 83, fObjlen = 15196, noutot = 0, nout=0, nin=8424, nbuf=15196
Warning in TBasket::ReadBasketBuffers: basket:fTracks.mPidElectron has fNevBuf=9, but fEntryOffset=0, trying to repair
Error in TBranchElement::GetBasket: File: /auto/rnc3/sarblyth/production62/Days/088//st_physics_5088046_raw_1030020.flowpicoevent
.root at byte:4556260, branch:(null), entry:150153100
(incomplete l-tree) R__unzip: error during decompression[/quote]

I think that this is related to the disk the files are on - the disk-vault is struggling to keep up and we know many people are using it. The files are fine and have been used under less stressed conditions without any issues. The code takes long but works without any problems on other files and at other times.

Is there any known issue with reading from NFS mounted volumes that are under intensive IO? Is there anything we can do to make ROOT more tolerant?

Sorry about the general questions - I would appreciate any advice regarding a better way to diagnose the issue.

Thanks,

Mark

2

You do not indicate which version you are using.
Could you try with version 4.01/04?

Rene

3

Hi Rene

Sorry about that, I am using v4-00-04. I can try to use v4-01-04 but its not currently
part of our framework so it will take a little while to try.

I’ll get back to you on it when I’ve tried it.

Thanks,

Mark