xrdcp: unable to mkdir by concurrently run xrdcp clients

[moscicki]

Version: 4.6.1-1

This is likely caused by two concurrent processes which are xrdcp to a destination path which does not exist. The first one does the mkdir to create the destination path and the other one tries to do the same but fails because the path in the meantime was already created by the first one.

Example:

2017-08-18 17:04:48,062 BackupThread env XRD_LOGLEVEL=Error XrdSecPROTOCOL=sss xrdcp --nopbar --verbose --force --retry 3 --tpc only root://localhost.localdomain//eos/?eos.lfn=fid:564441388&eos.ruid=2&eos.rgid=2 root://eosbackup.cern.ch//eos/backup/instance/user/data/170818/0000dc7c/564441388?eos.ruid=eosarchi&eos.rgid=c3&eos.app=backup_daily

2017-08-18 17:04:48,382 BackupThread ERROR: Transfer FAIL 564441388 spent 0.32 seconds XRDCP ERROR: '[2017-08-18 17:04:48.381064 +0200][Error ][Utility ] Unable to open target root://eosbackup.cern.ch:1094//eos/backup/instance/user/data/170818/0000dc7c/564441388?eos.app=backup_daily&eos.rgid=c3&eos.ruid=eosarchi&oss.asize=30019&tpc.key=0005cad34a05785059970210&tpc.lfn=/eos/&tpc.src=p06253937w63885.cern.ch:1095&tpc.stage=copy: [ERROR] Server responded with an error: [3006] Unable to mkdir /eos/backup/instance/user/data/170818/0000dc7c/; File exists\nRun: [ERROR] Server responded with an error: [3006] Unable to mkdir /eos/backup/instance/user/data/170818/0000dc7c/; File exists\n\n'

[xrootd-dev]

Generally, this is dependent on how this is done. If xrdcp is explicitly making directories at the destination then this problem may arrise and is consistent with POSIX behaviour. If the path is implicitly being created (i.e. the makepath on open has been specified) then this problem should not occur (it doesn't on non-eos systems). In this case, it's an eos issue. Andy

…

On Fri, 18 Aug 2017, moscicki wrote: Version: 4.6.1-1 This is likely caused by two concurrent processes which are xrdcp to a destination path which does not exist. The first one does the mkdir to create the destination path and the other one tries to do the same but fails because the path in the meantime was already created by the first one. Example: 2017-08-18 17:04:48,062 BackupThread env XRD_LOGLEVEL=Error XrdSecPROTOCOL=sss xrdcp --nopbar --verbose --force --retry 3 --tpc only root://localhost.localdomain//eos/\?eos.lfn=fid:564441388\&eos.ruid=2\&eos.rgid=2 root://eosbackup.cern.ch//eos/backup/instance/user/data/170818/0000dc7c/564441388\?eos.ruid=eosarchi\&eos.rgid=c3\&eos.app=backup_daily 2017-08-18 17:04:48,382 BackupThread ERROR: Transfer FAIL 564441388 spent 0.32 seconds XRDCP ERROR: '[2017-08-18 17:04:48.381064 +0200][Error ][Utility ] Unable to open target root://eosbackup.cern.ch:1094//eos/backup/instance/user/data/170818/0000dc7c/564441388?eos.app=backup_daily&eos.rgid=c3&eos.ruid=eosarchi&oss.asize=30019&tpc.key=0005cad34a05785059970210&tpc.lfn=/eos/&tpc.src=p06253937w63885.cern.ch:1095&tpc.stage=copy: [ERROR] Server responded with an error: [3006] Unable to mkdir /eos/backup/instance/user/data/170818/0000dc7c/; File exists\nRun: [ERROR] Server responded with an error: [3006] Unable to mkdir /eos/backup/instance/user/data/170818/0000dc7c/; File exists\n\n' -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: #568 ######################################################################## Use REPLY-ALL to reply to list To unsubscribe from the XROOTD-DEV list, click the following link: https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-DEV&A=1

[simonmichal]

The client is just setting the mkpath flag in the open request so this behaviour is server-side implementation dependent. So indeed, there is a good chance it is an eos problem.

Michal

[esindril]

This has been fixed on the EOS side.