before this, releases weren't deterministic and weren't even coordinated between image repos. This cleans it up, plus uses new docker ARGs to automate the release
Signed-off-by: Derek Perkins <derek@nozzle.io>
The redirect leads to a 404. However, adding a /debug/status to
the redirect URL does work. So, this is still an improvement.
I'll make a separate PR to append the /debug/status to the redirect
URL, which will fix this good. Then we'll have to build another helm
image.
Signed-off-by: Sugu Sougoumarane <ssougou@gmail.com>
This fixes orchestrator and pmm on k8s 1.16+. All of the other
statefulsets were already on apps/v1 so this just brings those two
components in line with the rest of the chart.
I have also fixed the printed proxy urls to actually work
Signed-off-by: Carson Anderson <ca@carsonoid.net>
K8s 1.15 added extra validation to CRD structure when using
the v1beta1 api. The CRD has been restructured to conform
to the new rules.
Signed-off-by: Carson Anderson <ca@carsonoid.net>
This is a backwards-compatible change for anyone
that has already installed the charts.
---
This will allow the CRD to be used in K8s clusters 1.7 or newer.
But some of the features, like the additionalPrinterColumns, will
only be available in 1.11 or newer, which is the minimum recommend
version
Signed-off-by: Carson Anderson <ca@carsonoid.net>
* Changed commands to work with latest minikube.
* Fixed 308 script to actually delete the job.
* Fixed minikube commands to work with the latest version.
Signed-off-by: Sugu Sougoumarane <ssougou@gmail.com>
there was a mixup in the orchestrator release process, hence the inconsistent binary and path matching for 3.1.0 and 3.1.1
Signed-off-by: Derek Perkins <derek@derekperkins.com>
Three major changes.
1. Turn off DiscoverByShowSlaveHosts - Because Vitess will reparent
during normal operations this causes Orchestrator to detect clusters
as Co-Master or Intermediate, this causes recovery not to work
2. Delay Mater Promotion - Don't blanked fail to promote if SQL Thread
is not up to date, just delay it instead
3. Put TER into a loop - If TER fails for any reason Vitess won't be
able to recover. We need that to succeed so we try it up to 10 times.
Also Add in the Datacenter Detect query, this doesn't effect
operations, but it does allow orchestrator to detect the Cell correctly
Signed-off-by: Dan Kozlowski <koz@planetscale.com>
Currently, when we access the ***vitess.io* with **HTTP**, it is
redirected to **HTTPS** automatically. So this commit aims to
replace **http://vitess.io** by **https://vitess.io** for security.
Co-Authored-By: Nguyen Phuong An <AnNP@vn.fujitsu.com>
Signed-off-by: Kim Bao Long <longkb@vn.fujitsu.com>
Shlomi Noach
Kewei Shang
Derek Perkins
Sugu Sougoumarane
Carson Anderson
Derek Perkins
Yuvraj
Gary Edgar
Morgan Tocker
Rohit Nayak
Ajeet jain
Abdullah Almariah
Connor Barnhill
Daniel Kozlowski
Eli Nudler
ch3rub1m
Kim Bao Long
huynq0911