Download Latest Version v0.1.5 source code.zip (604.9 kB) Google Add to Preferred Sources
Home / v0.1.4
Name Modified Size InfoDownloads / Week
Parent folder
checksums.txt 2026-09-03 957 Bytes
core_Linux_x86_64.tar.gz 2026-09-03 16.7 MB
core_Linux_arm64.tar.gz 2026-09-03 14.9 MB
core_0.1.4_Linux_arm64_debug.tar.gz 2026-09-03 30.1 MB
core_0.1.4_Linux_x86_64_debug.tar.gz 2026-09-03 32.9 MB
core_Darwin_arm64.tar.gz 2026-09-03 15.6 MB
core_Darwin_x86_64.tar.gz 2026-09-03 17.0 MB
core_0.1.4_Darwin_x86_64.tar.gz 2026-09-03 17.0 MB
core_0.1.4_Linux_arm64.tar.gz 2026-09-03 14.9 MB
core_0.1.4_Linux_x86_64.tar.gz 2026-09-03 16.7 MB
core_0.1.4_Darwin_arm64.tar.gz 2026-09-03 15.6 MB
README.md 2026-09-03 2.6 kB
v0.1.4 source code.tar.gz 2026-09-03 419.4 kB
v0.1.4 source code.zip 2026-09-03 606.9 kB
Totals: 14 Items   192.3 MB 0

Changelog

  • [17e09a] feat: plugin protocol round: verbs handshake, batched get-metrics, split streams (#708)
  • [98c91b] fix: bound every ssh channel open by the caller's context
  • [03e77a] fix: bring the redis store to etcd's contract on create, status and prefix reads (#706)
  • [965a49] fix: build the registry resolver per image call again
  • [161eb7] fix: close the store and rpc audit items
  • [f8a4f8] fix: collapse concurrent oras probes, tell a vm that has not booted from a verb it lacks, and move the cfs period next to its users
  • [1026ce] fix: drop the conflict markers left by the plugin protocol merge (#710)
  • [f47837] fix: drop the status read helper the lease merges left unused (#711)
  • [e6a5cd] fix: hand a new service watcher the latest addresses at once
  • [0f1ce4] fix: keep service discovery moving past a stuck subscriber and give each eru watch its own connection (#707)
  • [e1969e] fix: lease a workload status whose record is absent (#700)
  • [457dd7] fix: let a leadership sweep trust the listing's liveness, and admit heap entries where they are pushed
  • [534073] fix: let every calcium producer give up when its caller is gone
  • [8b30c4] fix: release containerd attach relays at task end
  • [b13e59] fix: return the write error when a cpumem node repair cannot be stored
  • [40e201] fix: revoke the lease a lost selfmon election granted (#699)
  • [d53215] fix: take pod locks in key order, and bound the selfmon fan-outs (#703)
  • [5695ce] perf: find the best cpu plan split by binary search and plan node capacities in parallel
  • [14d68c] perf: give long-lived ssh sessions their own connections, and stop holding a control slot to wait on a process
  • [092f35] perf: keep the request path off sentry, and fan a scrape out over nodes (#702)
  • [755775] perf: lock a node for the resource write, the pod only for a multi-node plan
  • [7103a7] perf: ping a node with an ssh keepalive instead of a session (#701)
  • [41500b] perf: pool etcd lock sessions across lock cycles (#698)
  • [73a00e] perf: renew an unchanged status lease instead of granting a new one (#704)
  • [553c8f] refactor: cpumem keeps its node records in the cluster store (#705)
  • [58e86f] refactor: drive remove and dissociate through one release path, and remap by node name
  • [0206d5] refactor: make the mock engine a plain struct outside the mocks tree
  • [a8d199] refactor: use the shared orphan status ttl in the etcd store
  • [095b14] review: name the parsed node resources cpumem hands around
  • [11a0a2] review: whole-repo round on the cluster, store, lock and rpc layers
  • [0c9d95] review: whole-repo round on the engines and resource plugins
Source: README.md, updated 2026-09-03