The syscall package is locked down and the comment in [1] advises to
switch code to use the corresponding package from golang.org/x/sys. Do
so and replace usage of package syscall with package
golang.org/x/sys/unix where applicable.
[1] https://github.com/golang/go/blob/master/src/syscall/syscall.go#L21-L24
This will also allow to get updates and fixes for syscall wrappers
without having to use a new go version.
Errno, Signal and SysProcAttr aren't changed as they haven't been
implemented in /x/sys/. Stat_t from syscall is used if standard library
packages (e.g. os) require it. syscall.SIGTERM is used for
cross-platform files.
This PR fixes the following issues:
1. Use ResourceStorageScratch instead of ResourceStorage API to represent
local storage capacity
2. In eviction manager, use container manager instead of node provider
(kubelet) to retrieve the node capacity and reserved resources. Node
provider (kubelet) has a feature gate so that storagescratch information
may not be exposed if feature gate is not set. On the other hand,
container manager has all the capacity and allocatable resource
information.
This PR adds two features:
1. add support for isolating the emptyDir volume use. If user
sets a size limit for emptyDir volume, kubelet's eviction manager
monitors its usage
and evict the pod if the usage exceeds the limit.
2. add support for isolating the local storage for container overlay. If
the container's overly usage exceeds the limit defined in container
spec, eviction manager will evict the pod.
This PR adds the check for local storage request when admitting pods. If
the local storage request exceeds the available resource, pod will be
rejected.
This PR adds the support for allocatable local storage (scratch space).
This feature is only for root file system which is shared by kubernetes
componenets, users' containers and/or images. User could use
--kube-reserved flag to reserve the storage for kube system components.
If the allocatable storage for user's pods is used up, some pods will be
evicted to free the storage resource.
Automatic merge from submit-queue (batch tested with PRs 45488, 45335, 45909, 46074, 46615)
Fix screwed-up log message format
It had two %-verbs and three arguments
**What this PR does / why we need it**:
Fixes kubelet log lines like this:
May 08 11:49:04 brya-1 kubelet[23248]: W0508 11:49:04.248123 23248 eviction_manager.go:128] Failed to admit pod kube-proxy-g3hjs_kube-system(55c1fbbb-33e4-11e7-b83c-42010a800002) - node has conditions: %v%!(EXTRA []v1.NodeConditionType=[MemoryPressure])
to remove the `%v%!(EXTRA`
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue (batch tested with PRs 46501, 45944, 46473)
fix func comment in helpers.go
**What this PR does / why we need it**:
fix func comment in helpers.go
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #
NONE
**Special notes for your reviewer**:
NONE
**Release note**:
```release-note
NONE
```
Automatic merge from submit-queue (batch tested with PRs 46022, 46055, 45308, 46209, 43590)
Eviction does not evict unless the previous pod has been cleaned up
Addresses #43166
This PR makes two main changes:
First, it makes the eviction loop re-trigger immediately if there may still be pressure. This way, if we already waited 10 seconds to delete a pod, we dont need to wait another 10 seconds for the next synchronize call.
Second, it waits for the pod to be cleaned up (including volumes, cgroups, etc), before moving on to the next synchronize call. It has a timeout for this operation currently set to 30 seconds.
Automatic merge from submit-queue
Moved qos to api.helpers.
**What this PR does / why we need it**:
The `GetPodQoS` is also used by other components, e.g. kube-scheduler and it's not bound to kubelet; moved it to api helpers so client-go.
**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #N/A
**Release note**:
```release-note-none
```
Automatic merge from submit-queue (batch tested with PRs 42369, 42375, 42397, 42435, 42455)
[Bug Fix]: Avoid evicting more pods than necessary by adding Timestamps for fsstats and ignoring stale stats
Continuation of #33121. Credit for most of this goes to @sjenning. I added volume fs timestamps.
**why is this a bug**
This PR attempts to fix part of https://github.com/kubernetes/kubernetes/issues/31362 which results in multiple pods getting evicted unnecessarily whenever the node runs into resource pressure. This PR reduces the chances of such disruptions by avoiding reacting to old/stale metrics.
Without this PR, kubernetes nodes under resource pressure will cause unnecessary disruptions to user workloads.
This PR will also help deflake a node e2e test suite.
The eviction manager currently avoids evicting pods if metrics are old. However, timestamp data is not available for filesystem data, and this causes lots of extra evictions.
See the [inode eviction test flakes](https://k8s-testgrid.appspot.com/google-node#kubelet-flaky-gce-e2e) for examples.
This should probably be treated as a bugfix, as it should help mitigate extra evictions.
cc: @kubernetes/sig-storage-pr-reviews @kubernetes/sig-node-pr-reviews @vishh @derekwaynecarr @sjenning
Automatic merge from submit-queue (batch tested with PRs 41074, 41147, 40854, 41167, 40045)
Add debug logging to eviction manager
**What this PR does / why we need it**:
This PR adds debug logging to eviction manager.
We need it to help users understand when/why eviction manager is/is not making decisions to support information gathering during support.
Fix threshold notifier build tags so the linux version is only built if cgo is
enabled, and the unsupported version is built if it's either not linux or not
cgo.
Automatic merge from submit-queue
kubelet: eviction: add memcg threshold notifier to improve eviction responsiveness
This PR adds the ability for the eviction code to get immediate notification from the kernel when the available memory in the root cgroup falls below a user defined threshold, controlled by setting the `memory.available` siginal with the `--eviction-hard` flag.
This PR by itself, doesn't change anything as the frequency at which new stats can be obtained is currently controlled by the cadvisor housekeeping interval. That being the case, the call to `synchronize()` by the notification loop will very likely get stale stats and not act any more quickly than it does now.
However, whenever cadvisor does get on-demand stat gathering ability, this will improve eviction responsiveness by getting async notification of the root cgroup memory state rather than relying on polling cadvisor.
@vishh @derekwaynecarr @kubernetes/rh-cluster-infra
Automatic merge from submit-queue
Cleanup kubelet eviction manager tests
It cleans up kubelet eviction manager tests
Extracted parts of tests that were similar to each other to functions
Automatic merge from submit-queue
kubelet: eviction: avoid duplicate action on stale stats
Currently, the eviction code can be overly aggressive when synchronize() is called two (or more) times before a particular stat has been recollected by cadvisor. The eviction manager will take additional action based on information for which it has already taken actions.
This PR provides a method for the eviction manager to track the timestamp of the last obversation and not take action if the stat has not been updated since the last time synchronize() was run.
@derekwaynecarr @vishh @kubernetes/rh-cluster-infra
Automatic merge from submit-queue
kubelet: eviction: allow minimum reclaim as percentage
Fixes#33354
xref #32537
**Release note**:
```release-note
The kubelet --eviction-minimum-reclaim option can now take precentages as well as absolute values for resources quantities
```
@derekwaynecarr @vishh @mtaufen
Automatic merge from submit-queue
Set imagefs rank and reclaim functions when nodefs+imagefs share comm…
Fixes#31192
I decided that the behavior should match the current output of the kubelet summary API. With no dedicated imagefs, the ranking and reclaim functions will match the nodefs ranking and reclaim functions.
/cc @ronnielai @vishh
Automatic merge from submit-queue
Eviction manager needs to start as runtime dependent module
To support disk eviction, the eviction manager needs to know if there is a dedicated device for the imagefs. In order to know that information, we need to start the eviction manager after cadvisor. This refactors the location eviction manager is started.
/cc @kubernetes/sig-node @kubernetes/rh-cluster-infra @vishh @ronnielai
Automatic merge from submit-queue
Add eviction-pressure-transitition-period flag to kubelet
This PR does the following:
* add the new flag to control how often a node will go out of memory pressure or disk pressure conditions see: https://github.com/kubernetes/kubernetes/pull/25282
* pass an `eviction.Config` into `kubelet` so we can group config
/cc @vishh
