Releases: TheNewNormal/corectl
Releases · TheNewNormal/corectl
v0.7.18
- user facing changes since 0.7.17
- fixed UI typo [https://github.com/TheNewNormal/corectl/commit/487abb36e13e9b44b37541afe7c843449f6aa4eb] / #96
- internal changes since 0.7.17
- refreshed dependencies and move from godep to govendor as external golang management tool [https://github.com/TheNewNormal/corectl/commit/425ef4c5cad652efa509799c2dc5872be66697c6] [https://github.com/TheNewNormal/corectl/commit/a2516f6e3978e018e9dc88b87f4cb278f2b730bf]
- embedded hyperkit updates [https://github.com/TheNewNormal/corectl/commit/da08d07562d89f1a51570e1f690a2385391aec4d] [https://github.com/TheNewNormal/corectl/commit/b82c5f2fab3e12bebeced4133ea0b23db94e68c2]
v0.7.17
- user facing changes since 0.7.15
- add a 'panic' button (f6d3a46)
- in the real world things just fail often unexpectedly and due to
reasons that one will never fully understand. In order to cope
with this circunstances and be fully prepared for them - because
they will happen corectl now has a "panic" option that will hard
kill a given running VM (or a randomly picked one). This feature
is intended as a practical way to test and reproduce both cluster
failure scenarios and the layouts resilient to them - en passant the stock the 'halt' option now complains when tasked
to halt stuff that doesn't exist
- reduced deliverables footprint (feea891)
- corectld is now just a symlink to corectl
- refactor for easier consumption of persistent root volumes (bebdb55)
- assuming a volume 'foo.qcow2' exists (created with 'qcow-tool
create --size=${SomeInteger}GiB foo.qcow2') one can now, on boot,
get it formated and partitioned via ignition when using the '-F'
flag before consuming it as the root volume.
(e.g. 'corectl run --name foo -r foo.qcow2 -F')
- assuming a volume 'foo.qcow2' exists (created with 'qcow-tool
- better handling and integration with hosts' DNS setup (632c249)
- on 816ebd2 we used macOS' stock facilities to expose to it our
embedded name server regardless of the host's setup. We now go one
step further - instead of running our embedded name server on host's
port 53 we run it on port 15353 and transparently port-forward via
pf it as :53 to the VMs. this should handle most, if not all, of the
"corner" cases that our user base has been hitting lately...
- on 816ebd2 we used macOS' stock facilities to expose to it our
- macOS Sierra compatibility
- internal changes since 0.7.15
- refreshed dependencies (8a398fb, c664855, ee9b741, ad81100)
- embeded hyperkit updates (e1af0b2, a400c78)
- Fix embedded etcd peer listen urls to use IPs (6d65b5c)
- cf. github.com/etcd-io/etcd/issues/6336
consumed via github.com/etcd-io/etcd/pull/6365
- cf. github.com/etcd-io/etcd/issues/6336
v0.7.15
- user facing changes since 0.7.14
- consume moby/hyperkit#49 - a (more) macOS Sierra friendly hyperkit (9a258b8)
- improvements to (VM's) shutdown logic (2e040f4)
- when shutting down
corectld
or halting multiple VMs those will
always be halted in reverse boot order - when hit by a signal
corectld
will now properly halt running VMs
- when shutting down
- internal changes since 0.7.14
v0.7.14
- user facing changes since 0.7.13
- add more resilience to NFS support (482b4f9)
- 'corectld' will no longer fail to initialize if it is unable to complete its NFS setup. Instead a warning will be shown by 'corectl' when a VM is launched with -H flag in that context.
- embedded DNS fixes
- reverse dns lookups are working again (24f4063)
- tune TTLs and add an NS record to embedded dns setup (10d3b6d)
TTLs were lowered a whole lot, as otherwise when interacting from the host one could stil be getting older IPs that hadn't expired yet for a given hostname that had been (re)booted several times in a short period.
- add more resilience to NFS support (482b4f9)
- internal changes since 0.7.13
- updated dependencies (cf21f04)
v0.7.13
- user facing changes since 0.7.12
- fix cloud-init behaviour (750d31b)
- the corectl hooks being passed though the default ignition
template weren't being passed exactly in the desired order
which made some features to not work as expected.
In particular$private_ipv4
and$public_ipv4
were not being
recognized properly in user provided cloud-configs since
/etc/environment
was being setup too late.
The etcd cluster example was updated to reflect this. - while working in the above issue another related surfaced
with the fix - let’s say an user is scripting and passes a
templated cloud-config to corectl and then progmamatically
deletes it from its local fs after corectl ends booting.
corectl would exit before actually have fetching that
cloud-config file (and when it would do it then it would
be gone). Now, and to cover this precise corner case, if
there's an user-provided cloud-config it is fetched and
saved early, at the same time as the ignition one and
saved for later use.
- the corectl hooks being passed though the default ignition
- fix cloud-init behaviour (750d31b)
- internal changes since 0.7.12
v0.7.12
- user facing changes since 0.7.11
- add support for QCow2 volumes …
- it's now the default / expected disk backend format
- closes #42
- ship tooling to ease QCw2 volumes creation (3e6018b)
qcow-tool
is now bundled together in order to provide our user base with a simple and
dependency-less way to create cqow2 volumes natively on macOS
- refactor embedded etcd and dns server logic (0320051)
- etcd and skydns are now embedded in "corectld" directly and
no longer are consumed as separate and external binaries - it is now possible to set, when starting "corectld", both the
internal's domain name (defaults to "coreos.local" and also
the recursive nameservers to use as fallback, default is to
use the stock Google ones - "8.8.8.8:53,8.8.4.4:53")
- add support for QCow2 volumes …
- internal changes since 0.7.11
- updated dependencies (84cceab)
v0.7.11
- user facing changes since 0.7.10
- better error handling on server startup (ce6ecc6)
- detects if stale
corectld.store
orcorectld.nameserver
somehow
were left running from previous sessions and if found kill them
and continue initialization instead of just popping an error and
exiting - exit with a much more conscise error message if
corectld.store
or
corectld.nameserver
ports are found binded to some foreign app
- detects if stale
- better error handling on server startup (ce6ecc6)
- no internal changes since 0.7.10
v0.7.10
- user facing changes (since 0.7.9)
- harden (server) shutdown logic (4ebe5a6)
- we now gracefully handle the corner cases where the bundled etcd or skyds went
away for reasons beyond our control
- we now gracefully handle the corner cases where the bundled etcd or skyds went
- allow a glimpse of templating on user-provided cloud-configs (91e3695)
- this feature is only supported from cloud-configs originally stored on the host filesystem
- initially only one variable is recognized:
- 'vm.Name' which is replaced by the VM's name
- this feature is only supported from cloud-configs originally stored on the host filesystem
- fix cluster example so that it works when not using git checkout builds (604efc9)
- fix
~/.coreos/running
permission grants (0a5c1b3)
- harden (server) shutdown logic (4ebe5a6)
- internal changes (since 0.7.9)
- remove dupe in signal handling (836a2f4)
v0.7.9
- user facing changes (since 0.7.8)
- the bundled
hyperkit
now comes with Dtrace support - complete fix for #80
- the bundled
- internal changes (since 0.7.8)
- refreshed dependencies
v0.7.8
- user facing changes (since v0.7.6)
- fix for (
corectld
) crasher bug introduced on 7f6b5bd when halting VMs
if you had hit this please kill manually all processes left running that start bycorectld
and then restartcorectld
- workarounds #80
when/Library/Preferences/SystemConfiguration/com.apple.vmnet.plist
is not yet
available in the local filesystem (because Internet Sharing wasn't used or
activated yet, or whatever) we'd just assume that the macOS defaults will be
used instead of failling inconditionally - closes #82 - insufficient permissions in ~/.coreos
- fix for (
- internal changes
- fix embedded etcd build inside sandboxes