You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: keps/sig-node/2400-node-swap/README.md
+5-3Lines changed: 5 additions & 3 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -211,7 +211,8 @@ This section is a recommendation for how to set up your nodes with swap if using
211
211
##### Disable swap for system critical daemon
212
212
213
213
As we were testing this feature, we found degration of services if you allow system critical daemons to swap.
214
-
The recommendation is to set the cgroup for the system slice to avoid swap (ie `memory.swap.max 0`).
214
+
This could mean that kubelet is performing slower than normal so if you experience this,
215
+
we recommend setting the cgroup for the system slice to avoid swap (ie `memory.swap.max 0`).
215
216
While doing this, we found that it is still possible for workloads to impact critical daemons.
216
217
217
218
##### Protect system critical daemons for iolatency
@@ -223,15 +224,16 @@ See [io-control](https://facebookmicrosites.github.io/cgroup2/docs/io-controller
223
224
224
225
##### Control Plane Swap
225
226
226
-
We only recommend enabling swap for the worker nodes. The control plane contains mostly Guaranteed QoS Pods, so swap may be disabled mostly.
227
+
We only recommend enabling swap for the worker nodes. The control plane contains mostly Guaranteed QoS Pods, so swap may be disabled for the most part.
227
228
The main concern would be swapping in the critical services on the control plane which can cause a performance impact.
228
229
229
230
##### Use of a dedicated disk for swap
230
231
231
232
We recommend using a separate disk for your swap partition. We recommend the separate disk be [encrypted](#security-risk).
232
233
If swap is on a partition or the root filesystem, workloads can interfere with system processes needing to write to disk.
233
234
If they occupy the same disk, it's possible processes can overwhelm swap and throw off the I/O of kubelet/container runtime/systemd, which would affect other workloads.
234
-
Swap space is located on a disk so it is imperative to make sure your disk is fast enough.
235
+
See [#protect-system-critical-daemons-for-iolatency] for more details on that.
236
+
Swap space is located on a disk so it is imperative to make sure your disk is fast enough for your use cases.
0 commit comments