Details
-
Type:
Bug
-
Status: Closed (View Workflow)
-
Priority:
Critical
-
Resolution: Fixed
-
Affects Version/s: 6.02/08
-
Component/s: I/O
-
Labels:None
-
Environment:
lxplus (SLC6), set up using ATLAS release 20.7.3.8 (LCG 75root6), x86_64-slc6-gcc48-opt
-
Development:
Description
As part of the ATLAS DQ monitoring we create rather large files with histograms and TTrees to keep metadata. We occasionally have files where attempting to delete one of these TTrees creates a corrupt file; an example is in /afs/cern.ch/user/p/ponyisi/public/testcase.root . The following will corrupt the file:
root [2] TFile *_file1 = TFile::Open("testcase.root", "UPDATE")
|
(class TFile *) 0x4485250
|
root [3] gDirectory->cd("/run_284285/lowStat_LB1001-1020/HLT/MuonMon/Coverage")
|
(Bool_t) true
|
root [4] gDirectory->Delete("metadata;1")
|
root [5] _file1->Write()
|
(Int_t) 0
|
root [6] .q
|
as evidenced by
root -l testcase.root
|
root [0]
|
Attaching file testcase.root as _file0...
|
(class TFile *) 0x1d8dc10
|
root [1] gDirectory->cd("/run_284285/lowStat_LB1001-1020/HLT/MuonMon/Coverage")
|
SysError in <TFile::Seek>: cannot seek to position -1807292313 in file testcase.root, retpos=-1 (Invalid argument)
|
Error in <TDirectoryFile::ReadKeys>: reading illegal key, exiting after 0 keys
|
Error in <TDirectoryFile::cd>: Unknown directory MuonMon
|
(Bool_t) false
|
Are we doing something wrong? This is a real worry for our DQ operations.