)]}' { "log": [ { "commit": "886d2892d1717bc130cfa008742c06c29f7ff186", "tree": "e9d67752d55d52e948662521e96d83f343473dea", "parents": [ "0565ea3191d445bdaab647213d59ed3d7218630a" ], "author": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Tue Apr 13 16:39:39 2021 +0200" }, "committer": { "name": "Leopold Schabel", "email": "leo@nexantic.com", "time": "Tue Apr 13 19:51:08 2021 +0200" }, "message": "metropolis/test/launch: add InsecureKey\n\nCurrently, the node startup parameters are empty. Let\u0027s populate them with a development (\u0027insecure\u0027) key when started from the launch library.\n\nTest Plan: Future revision in stack will make use of this.\n\nX-Origin-Diff: phab/D754\nGitOrigin-RevId: 0cfa3c1d71911423ff169afc027edb768151de67\n" }, { "commit": "37050126ef89ec30cc677c272471debe55ec0d69", "tree": "c64a64a622ec1c3e1e72fc12a6d4252c0e803cc1", "parents": [ "2999427c182463840a339cf0e82885d8a3b6e79f" ], "author": { "name": "Lorenz Brun", "email": "lorenz@nexantic.com", "time": "Tue Mar 30 14:00:27 2021 +0200" }, "committer": { "name": "Leopold Schabel", "email": "leo@nexantic.com", "time": "Tue Apr 13 11:04:01 2021 +0200" }, "message": "Implement Block PVCs in our storage backend\n\nThis implements full support for Block PVCs in our Kubernetes storage backend.\nThe block PVCs are backed by files made available to the pods using loop devices and\nhave read-only and online expansion support.\n\nThis also requires a Kubernetes patch because they call losetup if block PVCs are used\nwith CSI to establish a form of lock on the backing block device. This lock is not\nexclusive and does absolutely nothing for our use case and could get very expensive\non dense machines so I removed it.\n\nTest Plan: Comes with E2E tests\n\nX-Origin-Diff: phab/D746\nGitOrigin-RevId: 430d3f445286c0d3498b2153df333a19f3fcab89\n" }, { "commit": "30167f5cf55829d38f9d480466d7b5742c62a5fc", "tree": "fd89a3bb8a1c08b10d870a6b185b2deffa131cac", "parents": [ "9956e72c6c0b4f6436dc9493bc213965ee0cc191" ], "author": { "name": "Lorenz Brun", "email": "lorenz@nexantic.com", "time": "Wed Mar 17 17:49:01 2021 +0100" }, "committer": { "name": "Leopold Schabel", "email": "leo@nexantic.com", "time": "Tue Apr 13 11:03:56 2021 +0200" }, "message": "Add VM infrastructure smoke test\n\nThis adds an E2E test which exercises the VM infrastructure (Kubernetes, KVM device plugin and QEMU).\nThis test should ensure that nobody breaks the core infrastructure Metropolis VMs rely on.\n\nTest Plan: This is a test\n\nX-Origin-Diff: phab/D740\nGitOrigin-RevId: ddf629725dfb664ace5a50efee9ed9442962d6f7\n" }, { "commit": "9956e72c6c0b4f6436dc9493bc213965ee0cc191", "tree": "7842ac67432e3a187dda6a2dcb46d11088934159", "parents": [ "dca59d924dac4345099e5acd99405b5451d29cdb" ], "author": { "name": "Lorenz Brun", "email": "lorenz@nexantic.com", "time": "Wed Mar 24 18:48:55 2021 +0100" }, "committer": { "name": "Leopold Schabel", "email": "leo@nexantic.com", "time": "Tue Apr 13 11:03:53 2021 +0200" }, "message": "Add Loop Device package\n\nThis adds Loop device support in our Linux kernel and adds a Go package for working with them.\nIt also drive-by adds a pre-mounted tmpfs to ktest as that is quite useful in a lot of situations.\n\nTest Plan: Comes with ktests.\n\nX-Origin-Diff: phab/D745\nGitOrigin-RevId: fa06bcdddc033efb136f56da3b4a91159273bf88\n" }, { "commit": "0ed2f96a3a86aff2c9ce36289aa5d58a75f4d59b", "tree": "afbe1fb6cd0a1667e981edfe97969338437bdaca", "parents": [ "056042962060369bd7607ecfea51c515fc3a8140" ], "author": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Mon Mar 15 16:39:30 2021 +0100" }, "committer": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Mon Mar 15 16:39:30 2021 +0100" }, "message": "metropolis/proto: EnrolmentConfig -\u003e NodeParameters\n\nThis starts off the move to a node configuration API conforming to\nthe lifecycle management design document.\n\nInstead of an Enrolment Config used only to join an existing cluster, we\nmove to a NodeParameters proto that must always be given to a node if\nit\u0027s supposed to either bootstrap a new cluster or join an existing one.\n\nThis links the existing cluster management code (and its state machine)\nto work with this file. However, that state machine will be removed very\nsoon, anyway.\n\nWe also remove everything related to golden tickets.\n\nThis breaks multi-node tests.\n\nX-Origin-Diff: phab/D710\nGitOrigin-RevId: f22615fbccab975f2d5e6928bdc7387ab3aa5714\n" }, { "commit": "ddd6caff9edac56dad727a79eb5b0faf4dbd6cb9", "tree": "120710eb4a9acf0c3ad1086d9f6f6f3c850a0d70", "parents": [ "bcae658f9530e95cde2ac931beacae71c9fb240e" ], "author": { "name": "Lorenz Brun", "email": "lorenz@nexantic.com", "time": "Thu Mar 04 17:16:04 2021 +0100" }, "committer": { "name": "Lorenz Brun", "email": "lorenz@nexantic.com", "time": "Thu Mar 04 17:16:04 2021 +0100" }, "message": "Build mkfs.xfs using rules_cc\n\nThis drops the old big genrule for mkfs.xfs and replaces it with a nice rules_cc build system\nwith the help of bazel_cc_fix generated patches and our musl-based toolchain.\nWhile we\u0027re at it I bumped the versions of all related dependencies to their latest stable release.\nThis also means pulling in ini.h which is a dependency of the new xfstools version.\n\nInstructions to regenerate the patches are included in the spec files.\n\nToolchain selection is done by the existing transition in our rootfs rule so we automatically get a musl-built\nstatic binary when building for the rootfs.\n\nTest Plan: Tested with E2E tests, should fail fairly catastrophically if something were wrong.\n\nX-Origin-Diff: phab/D708\nGitOrigin-RevId: 648a05cdd08cfa84a8a9f4c057c52446e7005631\n" }, { "commit": "19eb0006edc79edc53fb53ea0eed67e93f4c8eba", "tree": "704a52ab75bde43409d80246cf23bce6b6be3467", "parents": [ "842536b10bd1b11e62317940feef215442a8ecb4" ], "author": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Thu Jan 21 14:25:25 2021 +0100" }, "committer": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Thu Jan 21 14:25:25 2021 +0100" }, "message": "third_party: bump Kubernetes to 1.19.7\n\nThis... didn\u0027t exactly go well. Turns out a change between rc.1 and rc.2\nbroke our runc runtime by enabling seccomp by default for pod sandboxes.\n\nWe work around this by reverting this change, and filing T916 to solve\nthis soon.\n\nThis fixes T910 and T909.\n\nTest Plan: kube bump, CI should run e2e, didn\u0027t run CTS.\n\nBug: T910, T909\n\nX-Origin-Diff: phab/D691\nGitOrigin-RevId: 78afca77c294895859e0af9150128d82677d875b\n" }, { "commit": "f12bedfa4cd144c3abc4deac58405067d55f9c87", "tree": "ddbc408e424a0ea8e446bcf0022ee16278202d63", "parents": [ "c3ad846e0eaf4cf008130a643ff247aa27531e17" ], "author": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Fri Jan 15 16:58:50 2021 +0100" }, "committer": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Fri Jan 15 16:58:50 2021 +0100" }, "message": "*: bump up Go dependencies\n\nThis started off as \u0027let\u0027s bump gVisor\u0027. However, pulling that thread\nresulted in quite a few things that also required bumping for the build\nto actually work. Here I come back from a day in the Bazel mines,\nbearing fruits of my labor.\n\nNotable changes:\n\n - bump up gVisor\n - bump up containerd\n - bump up Bazel\n - bump up rules_go, rules_docker, Gazelle\n - use google.golang.org/protobuf (the \u0027new\u0027 go proto package)\n - bump up gRPC (but not too much, as go-etcd is still straggling)\n\nNotable effects:\n\n - new gVisor supports TTY allocation (kubectl run -it\n --image\u003dubuntu:20.04 ubuntu bash now works!)\n\nNotable notes:\n\n - gVisor shim has new been rolled into the main gVisor package and is\n slightly easier to build (we can get rid of a bunch of patches).\n - Opencontainers\u0027 runtime-specs now follow containerd instead of gVisor\n - gVisor had to be taught to use the slightly newer runtime-specs via a\n new patch.\n - go_rule() in Starlark is now deprecated, and we had to change our\n Starlark rule definitions to use rule() instead. We also had to patch\n gVisor to do that (as there hasn\u0027t yet been a release that rolled\n this up).\n - Gazelle now supports different naming schemes for generated Go\n targets - either the old //foo/bar:go_default_library scheme, or a\n new and nicer //foo/bar:bar scheme. We currently force the usage of\n the old scheme, as switching over is probably not going to be easy\n (we use a lot of external Bazel files, and we have to wait for their\n compatibility with the new scheme first).\n - New Bazel/rules_go sets a TMPDIR long enough to generate paths (via\n ioutil.TempDir) to which sockets cannot be bound (108-byte limit).\n - The new protobuf API is incompatible with gogoproto. containerd/ttrpc\n uses gogoproto, but we are smart enough to pull in the old protobuf\n library as gogoproto\u0027s transitive dep. However, ttrpc also wants to\n use some proto-generated grpc bits, and that doesn\u0027t work. We have to\n pull in a ttrpc fork from a PR that hasn\u0027t yet been merged that fixes\n this issue.\n\nTest Plan: Refactor only, should be covered by tests.\n\nX-Origin-Diff: phab/D689\nGitOrigin-RevId: 1188c0605d25e7f40307fab5fd96e7019f3a9171\n" }, { "commit": "31370b07f0df2dc2765d812d4ce00a6b35185b16", "tree": "15563902eee9591083284441c8505b084b275d0a", "parents": [ "313816f41244d7520eb2b6f8c231328ee5b7a4ef" ], "author": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Thu Jan 07 16:31:14 2021 +0100" }, "committer": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Thu Jan 07 16:31:14 2021 +0100" }, "message": "*: git.monogon.dev -\u003e source.monogon.dev\n\nThis implements T882, setting our (virtual) GOPATH to source.monogon.dev\nfor this repository.\n\nTest Plan: Refactor, CI only.\n\nX-Origin-Diff: phab/D686\nGitOrigin-RevId: c5e2309089948ffc3a98e68e2e0e1cbb157d3a36\n" }, { "commit": "0be9be88224dd87eedb10436b11615fa59862271", "tree": "2cffcd0ca273ada48c0b42a36bd25bb1cc2da35c", "parents": [ "549b72b2d65051403301f53111509f77e88b379b" ], "author": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Thu Jan 07 15:23:44 2021 +0100" }, "committer": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Thu Jan 07 15:23:44 2021 +0100" }, "message": "metropolis: Lock down visibility rules\n\nThis formalizes the package structure introduced by D683.\n\nTest Plan: Pure refactor, CI only.\n\nX-Origin-Diff: phab/D684\nGitOrigin-RevId: 574aa14c71faf94f4a5c02a2110e2e3fef7d36ac\n" }, { "commit": "549b72b2d65051403301f53111509f77e88b379b", "tree": "b4e523d5d17e8130545e58b58870b4a18118a780", "parents": [ "696f39abb19ffcca03e9fc5a98681338216b1e7f" ], "author": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Thu Jan 07 14:54:19 2021 +0100" }, "committer": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Thu Jan 07 14:54:19 2021 +0100" }, "message": "metropolis: unify utility packages\n\nOne last sweeping rename / reshuffle.\n\nWe get rid of //metropolis/node/common and //golibs, unifying them into\na single //metropolis/pkg meta-package.\n\nThis is to be documented somwhere properly, but here\u0027s the new logic\nbehind selecting where to place a new library package:\n\n - if it\u0027s specific to k8s-on-metropolis, put it in\n //metropolis/node/kubernetes/*. This is a self-contained tree that\n other paths cannot import from.\n - if it\u0027s a big new subsystem of the metropolis core, put it in\n //metropolis/node/core. This can be imported by anything in\n //m/n (eg the Kubernetes code at //m/n/kubernetes\n - otherwise, treat it as generic library that\u0027s part of the metropolis\n project, and put it in //metropolis/pkg. This can be imported by\n anything within //metropolis.\n\nThis will be followed up by a diff that updates visibility rules.\n\nTest Plan: Pure refactor, CI only.\n\nX-Origin-Diff: phab/D683\nGitOrigin-RevId: 883e7f09a7d22d64e966d07bbe839454ed081c79\n" }, { "commit": "6df7c4f6b2c9a896357cb6c4e236d588f4e23277", "tree": "1f6690534fa123509708fe197ef3f35e6edf6903", "parents": [ "df952416e693f5b3180f1e69b6021a589cdc80d9" ], "author": { "name": "Lorenz Brun", "email": "lorenz@nexantic.com", "time": "Mon Dec 21 15:02:00 2020 +0100" }, "committer": { "name": "Lorenz Brun", "email": "lorenz@nexantic.com", "time": "Mon Dec 21 15:02:00 2020 +0100" }, "message": "Properly exit CTS on signals\n\nThis kills the loops when the global context is cancelled ensuring reliable termination\nof the CTS with Ctrl+C or a signal.\n\nTest Plan: Manually tested by aborting the CTS with Ctrl+C\n\nX-Origin-Diff: phab/D673\nGitOrigin-RevId: a2f367cd4a7a57bb573bd57656148681add048a2\n" }, { "commit": "df952416e693f5b3180f1e69b6021a589cdc80d9", "tree": "9ad8fdce45516e723b3ebe1aa49aec8b2d8194e6", "parents": [ "662b5b3119b0798980b887d1ef9fa1b5632aa7fb" ], "author": { "name": "Lorenz Brun", "email": "lorenz@nexantic.com", "time": "Mon Dec 21 14:59:36 2020 +0100" }, "committer": { "name": "Lorenz Brun", "email": "lorenz@nexantic.com", "time": "Mon Dec 21 14:59:36 2020 +0100" }, "message": "Fix nanoswitch logging after logtree refactor\n\nThe LogTree refactor left nanoswitch without a way to sink its logs as it doesn\u0027t have\na gRPC endpoint to read them from and logtree doesn\u0027t output to any stdout/stderr by default.\nThis adds a simple sink similar to what\u0027s currently used the node code.\n\nTest Plan: Manually tested\n\nX-Origin-Diff: phab/D672\nGitOrigin-RevId: 8698f9f920f36adf11fa1ef34a47723514eaa665\n" }, { "commit": "662b5b3119b0798980b887d1ef9fa1b5632aa7fb", "tree": "3e1fc4ab033530e6d579112ba500d2c6edb43368", "parents": [ "39f2f691726dc6e0a291aa8609085b835a313dad" ], "author": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Mon Dec 21 13:49:00 2020 +0100" }, "committer": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Mon Dec 21 13:49:00 2020 +0100" }, "message": "smalltown -\u003e metropolis\n\nThis pass removes all mentions of Smalltown, both from code and comments,\nand replaces them with appropriate new terminology.\n\nTest Plan: Refactor, covered by CI.\n\nX-Origin-Diff: phab/D674\nGitOrigin-RevId: 04a94d44ef07d46f7821530da5614daefe16d7ea\n" }, { "commit": "39f2f691726dc6e0a291aa8609085b835a313dad", "tree": "23d34844fa68be991d514bf8bd89bc5042779091", "parents": [ "686444ed962d75053832ce59527cf15a552a7fc0" ], "author": { "name": "Lorenz Brun", "email": "lorenz@nexantic.com", "time": "Mon Dec 21 14:51:42 2020 +0100" }, "committer": { "name": "Lorenz Brun", "email": "lorenz@nexantic.com", "time": "Mon Dec 21 14:51:42 2020 +0100" }, "message": "Make LaunchCluster not close DebugClient it\u0027s returning\n\nLaunchCluster currently closes the gRPC connection underpinning the DebugService\nit\u0027s returning. This causes all further calls to that service to immediately return with\na CANCELED error. There\u0027s no reason why it should do this (probably a refactoring artifact),\nso just remove the Close call. Fixes T881.\n\nTest Plan: CTS is not run as part of automated test still because it takes too long.\n\nBug: T881\n\nX-Origin-Diff: phab/D671\nGitOrigin-RevId: 6bfa382cba6a15b146b2f24311507456b58cdf98\n" }, { "commit": "686444ed962d75053832ce59527cf15a552a7fc0", "tree": "6c522bf2b538dafe4e5c345c96a353418ecd0c41", "parents": [ "553ab2b7a7f96e38ff7e381dcbefc095ed3520df" ], "author": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Mon Dec 21 14:21:14 2020 +0100" }, "committer": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Mon Dec 21 14:21:14 2020 +0100" }, "message": "launch-multi2: split up logs from nodes into prefixed lines\n\nCurrently it\u0027s impossible to tell apart logs from each node.\n\nIn general, we should move this over the the debug logging API instead\nof relying on qemu stdout, but this will do for now.\n\nTest Plan: Shouldn\u0027t affect any tests, as we don\u0027t actually test multi-node setups. Truth be told, we should.\n\nX-Origin-Diff: phab/D670\nGitOrigin-RevId: 7b4e170e634096bc40432fbef0844d9924957182\n" }, { "commit": "553ab2b7a7f96e38ff7e381dcbefc095ed3520df", "tree": "d8ea94719a7cf18b330a9adf772749a465419cb5", "parents": [ "77cb6c5ec3acadf02ad5005dd751cfbf0ec1602f" ], "author": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Mon Dec 21 13:59:33 2020 +0100" }, "committer": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Mon Dec 21 13:59:33 2020 +0100" }, "message": "nanoswitch: document shorter lease time\n\nThis was bumped down from 12h in D656, but the comment did not get\nupdated.\n\nTest Plan: Comment change. No test needed.\n\nX-Origin-Diff: phab/D669\nGitOrigin-RevId: fbd6bf76844b4eec4a20d4a52b472e2e47fff537\n" }, { "commit": "77cb6c5ec3acadf02ad5005dd751cfbf0ec1602f", "tree": "7ddfcdf78c489a5d6fad7a20bd3580d803407450", "parents": [ "26d41999e0c71813648c16ad84bba810c3b9d593" ], "author": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Sat Dec 19 00:09:22 2020 +0100" }, "committer": { "name": "Serge Bazanski", "email": "serge@nexantic.com", "time": "Sat Dec 19 00:09:22 2020 +0100" }, "message": "core -\u003e metropolis\n\nSmalltown is now called Metropolis!\n\nThis is the first commit in a series of cleanup commits that prepare us\nfor an open source release. This one just some Bazel packages around to\nfollow a stricter directory layout.\n\nAll of Metropolis now lives in `//metropolis`.\n\nAll of Metropolis Node code now lives in `//metropolis/node`.\n\nAll of the main /init now lives in `//m/n/core`.\n\nAll of the Kubernetes functionality/glue now lives in `//m/n/kubernetes`.\n\nNext steps:\n - hunt down all references to Smalltown and replace them appropriately\n - narrow down visibility rules\n - document new code organization\n - move `//build/toolchain` to `//monogon/build/toolchain`\n - do another cleanup pass between `//golibs` and\n `//monogon/node/{core,common}`.\n - remove `//delta` and `//anubis`\n\nFixes T799.\n\nTest Plan: Just a very large refactor. CI should help us out here.\n\nBug: T799\n\nX-Origin-Diff: phab/D667\nGitOrigin-RevId: 6029b8d4edc42325d50042596b639e8b122d0ded\n" } ] }