summaryrefslogtreecommitdiff
AgeCommit message (Collapse)Author
2015-07-28Additional hint for case where run succeedsinserted_srcBernard Ogden
Change-Id: Ie51b7d1ce67cd13d5622355c4b91f717898f11d3
2015-07-28Halt targets on successful runBernard Ogden
Change-Id: I3111b37c5222ac1c53f2cc01d658351313379a40
2015-07-28Add hint for post-build failure investigationBernard Ogden
Change-Id: I82e5053458e0f7e33265985e2e45285c5c3339cb
2015-07-27Move early IP report to first scriptBernard Ogden
This is even earlier, and de-pollutes the main script a little. Change-Id: I6dd6276190cad4af1c17b1c1ec78ed0627133bc0
2015-07-27Remove hacking session ismsBernard Ogden
We keep the bit that computes the IP address, even though lava-network will tell us the same thing later - this is for debugging of early problems. Change-Id: I753801fcc4c051401d0aabb218bb7fd02d33324f
2015-07-27Create debug modeBernard Ogden
Trap exits so that, if BENCH_DEBUG is set, we do not exit. Change-Id: I0d508e8f77b6ece882913a876f2ebd5054301cb4
2015-07-24Disable multiplexing in example ssh runeBernard Ogden
Invalid ControlPath guarantees that we won't reuse any existing socket (though there really shouldn't be an existing socket). I'd leave off the ControlMaster=no, given the invalid control path. But if I do this then I can't see stdout from the ssh process, for some reason. Change-Id: I511d201f45daa69b7712f4d16b71dfdeceb71d89
2015-07-24Check error code of Benchmark.jobBernard Ogden
Change-Id: I6cdac421d99b6d64d1405f389d0638583ac3fe92
2015-07-24Replace infinite sleep with 'results ready' messageBernard Ogden
Change-Id: Id692113b0270e316f93b2434b0723cb56d155048
2015-07-24Remove LAVA-in-benchmarking-scripts env varBernard Ogden
LAVA has been removed from scripts, env var isn't doing anything Change-Id: I8893cec0c8307ab39088d4977414eec1cd56fbf1
2015-07-24Change variable name from target_base to target_typeBernard Ogden
Better maps to the LAVA concept Change-Id: I220635d04a92ac6126c5ac7cc0a73e93952a8527
2015-07-24Add/fix comments/spacing in target setup loopBernard Ogden
Change-Id: Ic2643a5ade757a2ad82c277742f128e61f9583ee
2015-07-24Add a hint to userBernard Ogden
Change-Id: Id6961ebb49cea8b840f897532ea125d1f203cf23
2015-07-24Remove debugging printfsBernard Ogden
Change-Id: Ic64aeb988eb7d7975fc6b407f9fb9b8c05d573bd
2015-07-23Update the way we tell abe where to find the sourceBernard Ogden
And move it next to the rest of the source-related activity. Change-Id: Id2b4a5f624db2ff8a0936e189ae24cdec16f1ebe
2015-07-23Move agent request to just before it is neededBernard Ogden
Keep the code together, hold the agent for the shortest time possible. Change-Id: I058c79cb02fe85da618f7f5d7507ecf5eab88c50
2015-07-23Get local copy of sourceBernard Ogden
This lets us drop the agent before we call into abe - unfortunately the way I've structured this makes it hard to drop the agent just after the source gets checked out, so we check it out while we have the agent, then abe will check it out again, locally. Local copy acquired just before target setup, so that all the agent- dependent code is in one place - good for maintenance and for holding the agent for the shortest possible time. Change-Id: Ibbf1416a7a522283a9141fd469e3582754df3d98
2015-07-23Comment to explain socket deletionBernard Ogden
Change-Id: I203d4cfcf4afaa579ae3f3bb0fc9c20e0d255164
2015-07-23Remove unneeded packagesBernard Ogden
Remaining packages are needed by abe, or the benchmarking scripts or, in the case of libgfortran, for the convenience of building x86 SPEC in an uncontrolled way. Change-Id: I3c4ed02f7e0b45fbd9f2bb079b55b8af44f00923
2015-07-23Override SSH_AUTH_SOCK only for subprocessBernard Ogden
This allows our 'own' agent to continue to provide authentication for communication with target, so we are only dependent on the forwarded key(s) for initial target setup. Change-Id: I23762a03f0a369c341e730a2d884ef95f3906675
2015-07-23Use more specific option for localhost connectionBernard Ogden
Change-Id: I650eec8e9b8cdf5a4da928bcac1ee3fdb5b30f4e
2015-07-23Put onetime key into agentBernard Ogden
Change-Id: I492cfb87a7b545e5640d4d09bdec39d6f1609746
2015-07-23Specify maindir for Benchmark.jobBernard Ogden
Change-Id: If0f9b9f3fbc2251d918c84f12a2c37c4c1c3d4ed
2015-07-22Create a services fileBernard Ogden
Change-Id: Iceef9a732683c8bfe4aa39fa00aaa64525743a0e
2015-07-22Remove needless infinite sleepsBernard Ogden
They were there for debugging, but I do it in the exit trap Change-Id: If677d272ecda279e0af4318048a61bd79438eb41
2015-07-22Remove use of LAVA_SSH_KEYFILEBernard Ogden
Change-Id: Ic6a4fc2bd0f4bdf2369cf3b0f111c20516d7f237
2015-07-22Get a socket to put our one-time key on targetsBernard Ogden
Change-Id: Idf32073d286e28103c4a3c8c88c0139834a35250
2015-07-22Wait for a private key rather than an agentBernard Ogden
Private key doesn't have to be kept up - this is fine so long as we use a one-time key, as is the plan for Jenkins. Change-Id: I22767bd6a92d66d28e9dbada644d746aeb41b928
2015-07-22Iterate through not-host, rather than targetBernard Ogden
Lets us work around a multinode limitation (cannot have different target types within a single role). Change-Id: I4482e078702cc34e9558619f5be30aabd6ecc4ad
2015-07-22Default to getting toolchain over httpsBernard Ogden
Change-Id: I0784d7e50847dd507b98c9c21011d729d6e28c0d
2015-07-22Wait for an agent, not sourceBernard Ogden
For now, assume that agent-to-get-source & agent-to-access-targets are the same. Sufficient for prototyping with fakebench, which doesn't need an agent-to-get-source. Change-Id: Ic926c23cfe650255d619648d4607aaa0206d7e29
2015-07-21Remove debugging lava multinode runesBernard Ogden
Change-Id: I704f400d1a5c84ddedee2ae96c3b303635c38c39
2015-07-21Generate a conf file, not a json fileBernard Ogden
Change-Id: Id7323bac87f7295803b8e07b9599ee15979cc682
2015-07-21Construct list of targets in whitespace-safe wayBernard Ogden
Unfortunately we then have to export it in a non-whitespace-safe way, but at least we're well-behaved within the boundary of this script. Change-Id: Id514729f74808e18e58a21e144277c08fb3c41c1
2015-07-21Make array construction friendly to trapping undefined variablesBernard Ogden
Change-Id: I61bb7da646a489dafb165f37798874e485e511fb
2015-07-21Declare array correctlyBernard Ogden
Had declared it as an associative array - want to initialize it as an empty array. Change-Id: I9f1b416b75107d344ff3b230d9b234336c060a96
2015-07-21Disable multinode debug outputBernard Ogden
Was corrupting data read with lava-network Change-Id: I17d8b404b5701a5a37012551cff590a1482875b2
2015-07-21Check for zero-length IP addressBernard Ogden
Change-Id: I02f585e57ede52a89c2bb8a5a902ea08a7b29277
2015-07-21Remove incorrect internal quotesBernard Ogden
Change-Id: Id1a769f5778201d114bc05ee73b67fc91527a193
2015-07-21wip: more debugging infoBernard Ogden
Change-Id: Ie0531f15b99f0d609d243dd0113d00457da51322
2015-07-21Try doing a broadcast before a collectBernard Ogden
Effectively the same as doing a lava-send before a lava-wait-all? Working on hypothesis that the role stuff is busted. Change-Id: I673c18630148985cd916716b0b04df5ed0bc127f
2015-07-21Add a send, for case where we wait-all without a roleBernard Ogden
Change-Id: I9633e05ac27a8e0cccae9af2540987de6c131894
2015-07-21wip: do not be role-specificBernard Ogden
Change-Id: Ib056fec3ea9d833b34d3c161ed1c028fed51c24b
2015-07-21wip: lots of checkingBernard Ogden
Change-Id: I306749dd669d85354aef91c23c12e6192a09b3a9
2015-07-21wip: do not wait for sourceBernard Ogden
Change-Id: I20e37c294c9e435da96538b95a6440f291f699b4
2015-07-20wip-latest experiment, call lava-group FIRSTBernard Ogden
Change-Id: Iea560f3b8be902bb64cbaee6245a06beb4e078a0
2015-07-20WIP: Different multinode tackBernard Ogden
Change-Id: I3fa527e643282fb0f94599fdb55f37a4adf4febf
2015-07-14WIP: Testing multinodeBernard Ogden
Change-Id: I827211a72bf103bbd4f15390b27d42b8aeb26989
2015-07-14Figure out targets at runtimeBernard Ogden
Change-Id: I719bc3831a5e0f664125fd9b13f24d15dcb874e2
2015-07-14Don't actually exit from exit trapBernard Ogden
Change-Id: I68e67a2da84f1eed4b2e711d6aed48432c6d998c
generated by cgit v1.2.3 (git 2.25.1) at 2024-05-02 19:25:52 +0000