Netlogger Format to log XRootD Transfers/Accesses

[juztas]

It would be beneficial to have xrootd logging improved and maybe use the format like gridftp [1]. In this case sites could filter out transfers from/to or was it transfer or partial read and also what is the return code, start time/end time, etc..

@bbockelm fyi

[1]
http://netlogger.lbl.gov/doc

DATE=20190825171337.391455 HOST=transfer-2.ultralight.org PROG=globus-gridftp-server NL.EVNT=FTP_INFO START=20190825171336.718219 USER=cmsuser FILE=/mnt/hadoop/store/temp/user/lprado.ebec82b11611038a27d0e84af28c86023041fded/TTToSemiLeptonic_TuneCP5_PSweights_13TeV-powheg-pythia8/tthbb_sl_94X_v1/190822_141133/0001/ntuples_JESAbsoluteScaleup_Tree_1318.root BUFFER=235392 BLOCK=1048576 NBYTES=15171584 VOLUME=/ STREAMS=1 STRIPES=1 DEST=[131.225.204.196] TYPE=RETR CODE=226 TASKID=none retrans=4
DATE=20190825171400.100718 HOST=transfer-2.ultralight.org PROG=globus-gridftp-server NL.EVNT=FTP_INFO START=20190825171241.934773 USER=phedex FILE=/mnt/hadoop/store/mc/RunIIFall17MiniAODv2/ST_FCNC-TH_Tleptonic_HToWWZZtautau_Ctphi_CP5_13TeV-mcatnlo-madspin-pythia8/MINIAODSIM/PU2017_12Apr2018_94X_mc2017_realistic_v14-v1/270000/3E1901CA-41B5-E911-BA06-0090FAA572E0.root BUFFER=332800 BLOCK=1048576 NBYTES=3656287425 VOLUME=/ STREAMS=1 STRIPES=1 DEST=[131.225.207.226] TYPE=STOR CODE=226 TASKID=none
DATE=20190825171428.979333 HOST=transfer-2.ultralight.org PROG=globus-gridftp-server NL.EVNT=FTP_INFO START=20190825171354.811325 USER=cmsuser FILE=/mnt/hadoop/store/temp/user/dwinterb.7b5457f8effb276e55b427b22fcd49d760da706b/Aug22_MC_102X_2018/GluGluToMaxmixHToTauTauPlusTwoJets_M125_13TeV_amcatnloFXFX_pythia8/crab_GluGluToMaxmixHToTauTauPlusTwoJets_M125_amcatnloFXFX/190823_084946/0000/EventTree_186.root BUFFER=234912 BLOCK=1048576 NBYTES=655815949 VOLUME=/ STREAMS=1 STRIPES=1 DEST=[2A0C:5BC0:C8:2:B226:28FF:FE1E:C8F0] TYPE=RETR CODE=226 TASKID=none retrans=29

[bbockelm]

@juztas - this would be easy to do for third party copy (since that's a separate module from the rest of xrootd). Would that be sufficient?

[juztas]

It would be a good start, but in the end I would like to see in all (xrootd/dav/http/newsomething). In this case it would also allow to separate exactly the protocol is used for communications

[juztas]

Hi, @bbockelm do you have news on this? As a beginning for TPC maybe?

[bbockelm]

Hi @juztas - I remain fairly overwhelmed when it comes to Xrootd tasks. Any chance I could get you to try your hand at a few of the logging changes?

I'm perfectly supportive of the approach, I just have been struggling to find time.

[bbockelm]

@juztas - nice quiet weekend here. Please see #1156

[juztas]

Awesome!

…

On Sun, 15 Mar 2020 at 12:30, Brian P Bockelman ***@***.***> wrote: @juztas <https://github.com/juztas> - nice quiet weekend here. Please see #1156 <#1156> — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1049 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAE53NVF3M62K57LMXBGCDDRHUUFZANCNFSM4IPJVKXA> .

-- Justas Balcas

[juztas]

Can we re-open the ticket? This fixed for TPC - but not for other accesses (remote reading, davs, https).

[bbockelm]

Hi @juztas - can you open a separate ticket? In terms of metadata, I'd prefer to only have a single work item attached to an issue instead of adding scope.

[juztas]

@bbockelm would it be possible to include also START (and also END) timestamp in each log line? Success/failure to have start timestamp of transfer and finish. Now it is spread between 2 lines:

210929 08:18:43 14110 TPC_PullRequest: event=PULL_START, local=/store/mc/d.root, remote=https://cmsdcadisk.fnal.gov/d.root, user=cmsprod; Starting a push request
...
210929 08:22:04 14110 TPC_PullRequest: event=TRANSFER_SUCCESS, local=/store/mc/d.root, remote=https://cmsdcadisk.fnal.gov/d.root, user=cmsprod, bytes_transferred=3514218087, tpc_status=200

Also - does it have a thread ID or something? That would also be very beneficial to have (just in case we get multiple transfers transferring same file - rare - but possible).

[bbockelm]

@juztas - that seems reasonable. Can you open a separate ticket for this improvement? Any chance you want to take a swing at it?

[abh3]

I think we also agreed that we would not use an obsolete format but something more modern like JSON.

…

On Tue, 5 Oct 2021, Brian P Bockelman wrote: @juztas - that seems reasonable. Can you open a separate ticket for this improvement? Any chance you want to take a swing at it? -- You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub: #1049 (comment)

[bbockelm]

Oh, I don't think we literally want to use the Netlogger format... I read it as making the logs a bit easier on humans to work through.

Anyhow, let's get a new ticket opened and we can bikeshed a bit there.

[juztas]

@bbockelm ok, I can look into that (Just need to find all the bits to set xrd dev machine). Will create issue later