June 2013 - swift-bugs - mailman3.cels.anl.gov

[Bug 746] New: Cobalt provider on Eureka only uses 8 cores in multi-node jobs
by bugzilla-daemon＠mcs.anl.gov 30 Aug '13

30 Aug '13

https://bugzilla.mcs.anl.gov/swift/show_bug.cgi?id=746 Summary: Cobalt provider on Eureka only uses 8 cores in multi-node jobs Product: Swift Version: 0.93 Platform: PC OS/Version: Mac OS Status: ASSIGNED Severity: normal Priority: P2 Component: Providers AssignedTo: davidk(a)ci.uchicago.edu ReportedBy: wilde(a)mcs.anl.gov CC: hategan(a)mcs.anl.gov, davidk(a)ci.uchicago.edu, wozniak(a)mcs.anl.gov On Eureka, when I request coaster slots with nodegranularity=2 and have 8 Cobalt jobs running (for a total of 16 8-core nodes), is that I never see more than 64 jobs active. My throttle should allow 128 jobs. It looks to me like, possibly, we are only starting a coaster worker on one of the two nodes. I.e., our "multinode" logic in the submit file may be broken. Here's my sites file: <pool handle="cobalt"> <execution provider="coaster" jobmanager="local:cobalt" url="eureka.alcf.anl.gov" /> <filesystem provider="local" />  <profile namespace="globus" key="project">MTCScienceApps</profile> <profile namespace="globus" key="queue">default</profile> <profile namespace="karajan" key="jobthrottle">1.28</profile> <profile namespace="karajan" key="initialScore">10000</profile> <profile namespace="globus" key="jobsPerNode">8</profile>   <profile namespace="globus" key="slots">8</profile> <profile namespace="globus" key="maxTime">7200</profile>  <profile namespace="globus" key="maxWallTime">00:10:00</profile> <profile namespace="globus" key="nodeGranularity">2</profile> <profile namespace="globus" key="maxNodes">2</profile> <workdirectory>/home/wilde/LZ/run012/swiftwork</workdirectory> </pool> -- Configure bugmail: https://bugzilla.mcs.anl.gov/swift/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 5

[Bug 1000] New: Swift picks different $HOME
by bugzilla-daemon＠mcs.anl.gov 28 Jul '13

28 Jul '13

https://bugzilla.mcs.anl.gov/swift/show_bug.cgi?id=1000 Summary: Swift picks different $HOME Product: Swift Version: 0.94 Platform: PC OS/Version: Linux Status: NEW Severity: minor Priority: P2 Component: General AssignedTo: hategan(a)mcs.anl.gov ReportedBy: ketan(a)mcs.anl.gov Filing this apparently unusual phenomena reported by user Ahmad Qamar. He happened two have two login names associated with on Fusion. While running Swift run with a relatively large number of simulation jobs, one or two will occasionally fail because Swift picks up the another, unused username and looks for .globus in the homedir corresponding to that name. Some diagnosis: The environment does not show any evidence of any variable pointing to the other user. The only place where it becomes apparent that the user has two login names is on the finger command. The two login scenario on Fusion has been resolved since and that error has not shown up as I hear from the user. I am trying to reproduce this on an EC2 node by mimicking the same scenario. Filing this now to record the bug in the system recalling the verbal conversation with user. -- Configure bugmail: https://bugzilla.mcs.anl.gov/swift/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 2

[Bug 1018] New: Make app definitions in tc.data optional
by bugzilla-daemon＠mcs.anl.gov 18 Jul '13

18 Jul '13

https://bugzilla.mcs.anl.gov/swift/show_bug.cgi?id=1018 Summary: Make app definitions in tc.data optional Product: Swift Version: trunk Platform: PC OS/Version: Mac OS Status: ASSIGNED Severity: normal Priority: P2 Component: General AssignedTo: hategan(a)mcs.anl.gov ReportedBy: davidk(a)ci.uchicago.edu Make app definitions in tc.data optional. For example: app (file o) cat (file i) { cat @i stdout=@o; } There is no definition for cat in tc.data here. The command "cat" would still run on the worker node and be assumed to be available via $PATH for every pool entry you have defined. You should also be able to define a full pathname to an app without a tc.data entry: app (file o) cat (file i) { /bin/cat @i stdout=@o; } For situations where an app is available on one site and not another, you can define things as you currently do in tc.data and force apps to only run on selected sites. -- Configure bugmail: https://bugzilla.mcs.anl.gov/swift/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 1

[Bug 1026] New: Concurrent mapper fails in a recursion loop.
by bugzilla-daemon＠mcs.anl.gov 18 Jul '13

18 Jul '13

https://bugzilla.mcs.anl.gov/swift/show_bug.cgi?id=1026 Summary: Concurrent mapper fails in a recursion loop. Product: Swift Version: 0.94 Platform: PC URL: http://ci.uchicago.edu/~yadunandb/reiteration.tar OS/Version: Linux Status: NEW Severity: normal Priority: P2 Component: General AssignedTo: hategan(a)mcs.anl.gov ReportedBy: yadunandb(a)ci.uchicago.edu I'm trying to implement a simple feedback pattern in swift. Here's what I have : An application foo( ), which feeds back its output till ome condition cond( ) returns true. cond( ) just checks if a counter in the file has reached 100. The work( ) app, increments the counter in the input file by 1. The following pseudo swift code should go to a 100 level deep recursion, but fails with could not create file error. (file output) foo (file input) { if ( input meets conditions ){ output = input; // tail end of recursion } else { // do some work on input and pass it back to foo ( ) file temp = work (input); output = foo ( temp ); } } >From what I understand here, the concurrent mapper is creating temporary files by appending the filename at each level of the recursion with some chars resulting in filenames exceeding the 255 char limit by the 50th level. Please check the URL attached for a test script which would reproduce this issue. -- Configure bugmail: https://bugzilla.mcs.anl.gov/swift/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 1

[Bug 1028] New: readData in a foreach loop fails to compile.
by bugzilla-daemon＠mcs.anl.gov 16 Jul '13

16 Jul '13

https://bugzilla.mcs.anl.gov/swift/show_bug.cgi?id=1028 Summary: readData in a foreach loop fails to compile. Product: Swift Version: 0.94 Platform: PC OS/Version: Linux Status: NEW Severity: normal Priority: P2 Component: General AssignedTo: hategan(a)mcs.anl.gov ReportedBy: yadunandb(a)ci.uchicago.edu Here's the piece of code that fails to compile : type file; app (file o) mysim () { random stdout=@filename(o); } int rand_num[]; foreach i in [0:9] { file f <single_file_mapper; file=@strcat("output/sim_",i,".out")>; f = mysim(); rand_num[i] = readData(f); } Here's the output I get when compiling the code: Could not start execution Failed to convert .swiftx to .kml for bug.swift no such attribute: partialClose in template context [callInternal] -- Configure bugmail: https://bugzilla.mcs.anl.gov/swift/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 1

[Bug 1022] New: ssh provider prompts incorrectly for supplied auth info
by bugzilla-daemon＠mcs.anl.gov 12 Jun '13

12 Jun '13

https://bugzilla.mcs.anl.gov/swift/show_bug.cgi?id=1022 Summary: ssh provider prompts incorrectly for supplied auth info Product: Swift Version: 0.94 Platform: All OS/Version: Linux Status: ASSIGNED Severity: normal Priority: P2 Component: Providers AssignedTo: hategan(a)mcs.anl.gov ReportedBy: wilde(a)mcs.anl.gov CC: hategan(a)mcs.anl.gov, davidk(a)ci.uchicago.edu, wozniak(a)mcs.anl.gov With an auth.defaults entry correctly coded for pathphrase auth, swift 0.94 works fine. However if you remove (just) the passphrase line from a site in auth.defaults (so that Swift will prompt for the passphrase) then you get the following un-desired behavior: Swift first prompts for the username and key file location, providing what looks like the settings that it got from auth.defaults, as if its using this as defaults and just wants a <return> for those lines. Then it prompts for the password. However, if you just type <return> for the username and keyfile prompt, it doesnt seem to use the values it presented in [] brackets as the defaults. Instead, you have to actually re-type what you were just prompted with. For example: [wilde] wilde [/home/wilde/.ssh/id_rsa-swift] /home/wilde/.ssh/id_rsa-swift as in: mid$ swift -config cf -tc.file apps -sites.file ssh.xml echo.swift Swift 0.94 swift-r6414 (swift modified locally) cog-r3648 RunID: 20130613-0419-hk1in4k3 Progress: time: Thu, 13 Jun 2013 04:19:38 +0000 [wilde] wilde [/home/wilde/.ssh/id_rsa-swift] /home/wilde/.ssh/id_rsa-swift [null] U[dgkrux&,/1;@DHNPRY\`dhy,8:=X Progress: time: Thu, 13 Jun 2013 04:20:08 +0000 Selecting site:10 Initializing site shared directory:1 Progress: time: Thu, 13 Jun 2013 04:20:11 +0000 Selecting site:7 Initializing site shared directory:1 Stage in:3 Progress: time: Thu, 13 Jun 2013 04:20:12 +0000 Submitting:4 Submitted:1 Active:6 Progress: time: Thu, 13 Jun 2013 04:20:13 +0000 Active:3 Checking status:1 Stage out:7 Progress: time: Thu, 13 Jun 2013 04:20:17 +0000 Stage out:10 Finished successfully:1 Progress: time: Thu, 13 Jun 2013 04:20:18 +0000 Stage out:3 Finished successfully:8 Final status: Thu, 13 Jun 2013 04:20:19 +0000 Finished successfully:11 mid$ There's two things that seem wrong here: 1) It seems wrong to be prompted at all for username and passphrase when these were supplied in auth.defaults 2) it seems that if Swift is trying to give you a chance to override the settings from auth.defaults, then it should at least handle the values it has as real deafults, and accept a <return> to let the uer use those values. -- Configure bugmail: https://bugzilla.mcs.anl.gov/swift/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 1

[Bug 1014] New: mpi jobs fail on stampede
by bugzilla-daemon＠mcs.anl.gov 11 Jun '13

11 Jun '13

https://bugzilla.mcs.anl.gov/swift/show_bug.cgi?id=1014 Summary: mpi jobs fail on stampede Product: Swift Version: 0.94 Platform: PC OS/Version: Linux Status: NEW Severity: normal Priority: P2 Component: General AssignedTo: davidk(a)ci.uchicago.edu ReportedBy: ketan(a)mcs.anl.gov MPI jobs fails with a signal 15 after 2-4 minutes in running. My submit host is communicado and swift modes tested are both ssh-cl:slurm and gt2:gt2:slurm. The logs, error message on Swift stdout, slurm job script, slurm.submit.stderr can be found here: http://ci.uchicago.edu/~ketan/stampede.mpi.logs.tgz -- Configure bugmail: https://bugzilla.mcs.anl.gov/swift/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 11

[Bug 1012] New: Swift quick start guide references 0.93
by bugzilla-daemon＠mcs.anl.gov 11 Jun '13

11 Jun '13

https://bugzilla.mcs.anl.gov/swift/show_bug.cgi?id=1012 Summary: Swift quick start guide references 0.93 Product: Swift Version: 0.94 Platform: All URL: http://www.ci.uchicago.edu/swift/guides/release-0.94/q uickstart/quickstart.html OS/Version: All Status: NEW Severity: normal Priority: P2 Component: Documentation AssignedTo: wozniak(a)mcs.anl.gov ReportedBy: skrieder(a)iit.edu The Swift/K quick start guide references swift-0.93. -- Configure bugmail: https://bugzilla.mcs.anl.gov/swift/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 1

[Bug 961] New: Wrapper staging file not found
by bugzilla-daemon＠mcs.anl.gov 11 Jun '13

11 Jun '13

https://bugzilla.mcs.anl.gov/swift/show_bug.cgi?id=961 Summary: Wrapper staging file not found Product: Swift Version: trunk Platform: PC OS/Version: Mac OS Status: NEW Severity: normal Priority: P2 Component: General AssignedTo: hategan(a)mcs.anl.gov ReportedBy: davidk(a)ci.uchicago.edu I just tried a quick test of wrapper staging in trunk, but am getting an error that the input file is not found. The error is: Caused by: Cannot stage in localhost/data.txt. File not found. The full log is at http://www.ci.uchicago.edu/~davidk/logs/catsn-20130305-1113-r1xsgfcc.log. I'm not sure if I have things set up correctly or not. Here is the setup I am testing with. cf -- wrapperlog.always.transfer=true sitedir.keep=true execution.retries=0 lazy.errors=true status.mode=provider use.provider.staging=false provider.staging.pin.swiftfiles=false use.wrapper.staging=true wrapper.staging.local.server=file:// sites.xml -------- <config> <pool handle="localhost"> <execution provider="local" url="none"/> <profile namespace="karajan" key="jobThrottle">0</profile> <profile namespace="karajan" key="initialScore">10000</profile> <filesystem provider="local"/> <workdirectory>/home/davidk/swiftwork</workdirectory> <scratch>/scratch/local/davidk</scratch> </pool> </config> catsn.swift --------- type file; app (file o) cat (file i) { cat @i stdout=@o; } file out[]<simple_mapper; location=".", prefix="catsn.",suffix=".out">; foreach j in [1:@toInt(@arg("n","10"))] { file data <"data.txt">; out[j] = cat(data); } -- Configure bugmail: https://bugzilla.mcs.anl.gov/swift/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 4

[Bug 1008] New: Errors in User Guide coaster section
by bugzilla-daemon＠mcs.anl.gov 07 Jun '13

07 Jun '13

https://bugzilla.mcs.anl.gov/swift/show_bug.cgi?id=1008 Summary: Errors in User Guide coaster section Product: Swift Version: 0.94 Platform: PC OS/Version: Mac OS Status: ASSIGNED Severity: major Priority: P1 Component: Documentation AssignedTo: ketan(a)mcs.anl.gov ReportedBy: wilde(a)mcs.anl.gov 18.5. Submitting to a specific job queue This section is not coaster-specific. It probably should not be a section. scheduler is misspelled "schedular" several times 18.6. Jobs and Coasters Parameters Equation "The parameters provided in the coasters specification affect jobs submitted by Swift to the scheduling system. Jobs Submitted = slots X jobsPerNode X nodeGranularity X maxNodes" This formula is both incorrect and ambiguous. By "jobs" I assume it meant "local resource manager jobs" (eg PBS jobs), not Swift app() tasks. And if thats the case, the formula is incorrect. The number of LRM jobs is "slots". And nodeGranularity and maxNodes do not have a multiplicative effect. Please revise this section. As stated it just makes the complex table of coaster attributes above even more confusing. -- Configure bugmail: https://bugzilla.mcs.anl.gov/swift/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 1