kubernetes: add mount of /tmp #137
Comments
|
@johnbelamaric, to clarify - you mean as temporary fix for people using 1.3.1? |
|
Yes. Especially since most people will use 1.3.1 for now due to backwards
incompatibilities in 1.4, correct?
…On Thu, Mar 28, 2019 at 1:24 PM chrisohaver ***@***.***> wrote:
to clarify - you mean as temporary fix for people using 1.3.1?
This issue is already fixed in 1.4.0.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#137 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AJB4s8ekvXVEs-bzNUF4sIi_G303p-SLks5vbSURgaJpZM4cREJD>
.
|
|
NOTE:
For people having an issue with glog/klog, solution could be: |
|
Oh, so 1.4.1 will be OK and won't run afoul of deprecation policy in K8s?
…On Thu, Mar 28, 2019 at 3:04 PM Francois Tur ***@***.***> wrote:
NOTE:
- 1.4.0 could be compatible for most of people - at least would not
reject a config -. But there is a bug on reload - which is used on
default config for Kubernetes
- any further version (currently master) hhas backward incompatibility
and the default config provided with k8s v1.14 would fail CoreDNS.
For people having an issue with glog/klog, solution could be:
1- mount /tmp (as proposed here)
2- either use the version 1.3.0
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#137 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AJB4s0eQJAgHw26R89DLmv03IkcN3uniks5vbTxQgaJpZM4cREJD>
.
|
|
Hum ... I am not sure I understand. So you cannot use an image CoreDNS 1.4.1 for a Corefile we ship in k8s v1.14 - or earlier. |
|
So, there will be no 1.4.0, we are going straight to 1.5.0? Because unless
we bump the minor version, we can't do the incompatible release.
We either need to release a 1.4.1 that fixes reload, or we need to add this
/tmp and stay at 1.3.1 in k8s until the deprecation period there has passed.
…On Thu, Mar 28, 2019 at 3:10 PM Francois Tur ***@***.***> wrote:
Hum ... I am not sure I understand.
1.4.1 does not exist yet ... let say we build it now, it will be out of
mater that is already incompatible (no proxy, and "upstream" option of
kubernetes does not exist anymore)
So you cannot use an image CoreDNS 1.4.1 for a Corefile we ship in k8s
v1.14 - or earlier.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#137 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AJB4sx3nXpdizbGegMKJCRuUlRPgRxsTks5vbT3XgaJpZM4cREJD>
.
|
|
You are right, I understand the plan is to go straight to 1.5.0 |
|
But 1.4.1 cannot get out of current |
|
emptyDir for glog should not be needed. What would be needed is a test that catches when glog/klog starts barfing output again. |
|
It's needed for 1.3.1, which doesn't have the fix. |
|
Yeah, i think this workaround should be in place until we update the image to a version that fixes the g/klog issue, at which point we should remove it. @rajansandeep, can you create a PR to add this to the Deployment yaml... see |
|
Perhaps we should be tagging releases in this repo. |
In order to fix coredns/coredns#2464 and related issues with crashing due to glog/klog issues, we should just add an EmptyDir volume and mount it on /tmp
The text was updated successfully, but these errors were encountered: