ADR030: Allowed Pod disruptions

  • Status: accepted

  • Deciders:

    • Felix Hennig

    • Lars Francke

    • Sascha Lautenschläger

    • Sebastian Bernauer

    • Sönke Liebau

  • Date: 2023-09-15

Context and problem statement

Downtime of products is always bad, but sometimes Pods need to be restarted to roll out updates or renew certificates. To prevent services from becoming unavailable we need to make sure that there is always a certain number of Pods still online when restarting Pods. Kubernetes has a concept called PodDisruptionBudget (PDB) to define the number of Pods that need to be kept online or the number of Pods that can safely be taken offline. We want to use this functionary to either prevent services outages entirely or try to keep them to a minimum. PDBs are defined not on a StatefulSet or Deployment, but with a selector over labels, so they can also span Pods from multiple StatefulSets.

Example use-cases

  1. As a user I want an HDFS and it (or parts) should not be disturbed by planned pod evictions (for example for a certificate renewal). I expect this to be the default behaviour.

  2. As a user I want to configure maxUnavailable on the role (e.g. datanode) across all rolegroups (e.g. dfs replicas 3 and only a single datanode is allowed to go down - regardless of the number of rolegroups), so that no datanode is a single point of failure. Similarly for ZooKeeper, I want to define PDBs at role level as ZK quorum is independent of role groups.

  3. As a user I want to override defaults to maybe have less availability but faster rollout times in rolling redeployments; for example a Trino cluster that could take more than 6 hours to rolling redeploy, as the graceful shutdown of Trino workers takes a considerable amount of time - depended on the queries getting executed.

  4. As a user I want to configure maxUnavailable on rolegroups individually, as I e.g. have some fast datanodes using SSDs and some slow datanodes using HDDs. I want to have always X number of fast datanodes online for performance reasons.

  5. As a user I want a Superset/NiFi/Kafka and they (or parts) should not be disturbed by planned pod evictions.

  6. As a user I might want to define PDBs across roles or on other specific Pod selections, in that case I want to be able to disable the Stackable generated PDBs.

We expect the majority of users to either use default PDB settings or define PDBs at a role level. Role group configuration like in use-case 4 has merit but seems like a more niche usage scenario.

Technical considerations

We have the following constraints:

If we use arbitrary workloads and arbitrary selectors (for example when selecting Pods from multiple StatefulSets) we have the following constraints: * only .spec.minAvailable can be used, not .spec.maxUnavailable. * only an integer value can be used with .spec.minAvailable, not a percentage.

This means that if we select any Pods that are not part of a StatefulSet or Deployment etc. then we are bound by these constraints. Preliminary testing showed that .spec.maxUnavailable works with multiple StatefulSets.

You can use a selector which selects a subset or superset of the pods belonging to a workload resource. The eviction API will disallow eviction of any pod covered by multiple PDBs, so most users will want to avoid overlapping selectors.

We need to create PDBs in such a way that every Pod is only selected once. This is easiest if a selector is defined per role or for all role groups individually. Excluding certain labels is also possible my using match expressions, but we did not test whether is conflicts with the first constraint about arbitrary selectors. To support the user creating their own custom PDBs we need to support disabling PDB generation to prevent overlapping selectors.

Decision drivers

  • Common use-cases should be easy to configure.

  • Principle of least surprise: CRD configuration settings and their interactions in case of multiple settings need to be easy to comprehend to prevent user error.

  • Extendable design, so that we can later non-breaking add new functionality, such as giving the chance to configure PDBs on roleGroup level as well.

  • Simple implementation (far less important)

Decision outcome

Option 1 was picked.

Considered options

Option 1

Introduce a new roleConfig at role level and put PDBs in there. Only role level PDBs are supported, for role group level the PDBs should be disabled and the user needs to create PDBs manually. The roleConfig is put in place to not put the PDB setting directly in the role.

spec:
  nameNodes:
    roleConfig: # <<<
      podDisruptionBudget: # optional
        enabled: true # optional, defaults to true
        maxUnavailable: 1 # optional, defaults to our "smart" calculation
    roleGroups:
      default:
        replicas: 2
  dataNodes:
    # use pdb defaults
    roleGroups:
      default:
        replicas: 2

Pros

  • simple to understand

  • covers the majority of use cases

  • still leaves the option to disable and roll your own

Cons

  • Yet another "config" (config, clusterConfig and now roleConfig as well)

    • That’s kind of the way the real world is: There are some thing you can configure on cluster level (e.g. ldap), role level (pdbs) and role group level (resources). This models this the closest.

  • Its not possible to define PDBs on rolegroups without the user deploying it’s own PDBs.

In the discussion the option of having the PDB directly in the role without a roleConfig was briefly discussed but not considered as an option due to being too messy, so it is not listed as an explicit option here.

Option 2 - PDB in config, but only at role level

Instead of inventing a new roleConfig setting, put the PDB in the config. This might seem better at first, but usually settings in config can also be set at role group level, and in this case, that would not be true.

spec:
  nameNodes:
    config: # <<<
      podDisruptionBudget:
        enabled: true
        maxUnavailable: 1
    roleGroups:
      default:
        replicas: 2
        config: {}
          # no such field as podDisruptionBudget

Pros

  • Everything configurable is below config, no new roleConfig

  • Like Option 1, covers configuration of the most important use cases

Cons

  • spec.nameNodes.config is not similar to spec.nameNodes.roleGroups.default.config ⇒ Confusing to the user

    • thinking more about it, it might be confusing that the setting is not "copied" to all role groups like other settings like resources or affinities.

  • Still no option to configure role group level PDBs

  • Possibly complicated to implement, due to config usually being identical at role and role group level

Option 3: PDB in config with elaborate merge mechanism

Similar to Option 2, the PDB setting is located in the config but it is actually possible to use it at both role and role group level. We develop a semantic merge mechanism that would prevent overlapping PDBs.

CRD Example 
apiVersion: hdfs.stackable.tech/v1alpha1
kind: HdfsCluster
metadata:
  name: simple-hdfs
spec:
  image:
    productVersion: 3.3.4
  clusterConfig:
    zookeeperConfigMapName: simple-hdfs-znode
  nameNodes:
    config:
      podDisruptionBudget:
        enabled: true
        maxUnavailable: 2
    roleGroups:
      hdd:
        replicas: 16
        config:
          podDisruptionBudget:
            maxUnavailable: 4
      ssd:
        replicas: 8
        config:
          podDisruptionBudget:
            enabled: false
      in-memory:
        replicas: 4

would end up with something like

apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: simple-hdfs-datanodes-hdds
spec:
  maxUnavailable: 4
  selector:
    matchLabels:
      app.kubernetes.io/name: hdfs
      app.kubernetes.io/instance: simple-hdfs
      app.kubernetes.io/component: datanode
      app.kubernetes.io/rolegroup: hdd
---
apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: simple-hdfs-datanodes-not-hdds
spec:
  maxUnavailable: 2
  selector:
    matchLabels:
      app.kubernetes.io/name: hdfs
      app.kubernetes.io/instance: simple-hdfs
      app.kubernetes.io/component: datanode
    matchExpressions:
      - key: app.kubernetes.io/rolegroup
        operator: NotIn
        values:
          - hdd
      - key: app.kubernetes.io/rolegroup
        operator: NotIn
        values:
          - in-memory

Pros

  • Fits into the existing config structure

  • Allows configuring role config level PDBs and even hybrid configs

Cons

  • Complex merge mechanism possibly difficult to understand and therefore easy to use the wrong way

  • Complex mechanism also not trivial to implement

Option 4 - PDB in config with normal "shared role group config" behaviour

Again we put the PDB in the config section but simply use the normal "copy" behaviour for this setting. This would be simple and easy to understand, but does not allow for true role level PDBs.

CRD Example 
spec:
  dataNodes:
    config:
      podDisruptionBudget:
        maxUnavailable: 2
    roleGroups:
      hdd:
        replicas: 16
      ssd:
        replicas: 8
      in-memory:
        replicas: 4

would end up with

apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: simple-hdfs-datanodes-hdds
spec:
  maxUnavailable: 2
  selector:
    matchLabels:
      app.kubernetes.io/name: hdfs
      app.kubernetes.io/instance: simple-hdfs
      app.kubernetes.io/component: datanode
      app.kubernetes.io/rolegroup: hdd
---
apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: simple-hdfs-datanodes-hdds
spec:
  maxUnavailable: 2
  selector:
    matchLabels:
      app.kubernetes.io/name: hdfs
      app.kubernetes.io/instance: simple-hdfs
      app.kubernetes.io/component: datanode
      app.kubernetes.io/rolegroup: ssd
---
apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: simple-hdfs-datanodes-hdds
spec:
  maxUnavailable: 2
  selector:
    matchLabels:
      app.kubernetes.io/name: hdfs
      app.kubernetes.io/instance: simple-hdfs
      app.kubernetes.io/component: datanode
      app.kubernetes.io/rolegroup: in-memory
spec:
  nameNodes:
    config:
      podDisruptionBudget:
        enabled: true
        maxUnavailable: 2
    roleGroups:
      hdd:
        replicas: 16
        config:
          podDisruptionBudget:
            maxUnavailable: 4
      ssd:
        replicas: 8
        config:
          podDisruptionBudget:
            enabled: false
      in-memory:
        replicas: 4

would end up with

apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: simple-hdfs-datanodes-hdds
spec:
  maxUnavailable: 4
  selector:
    matchLabels:
      app.kubernetes.io/name: hdfs
      app.kubernetes.io/instance: simple-hdfs
      app.kubernetes.io/component: datanode
      app.kubernetes.io/rolegroup: hdd
---
apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: simple-hdfs-datanodes-hdds
spec:
  maxUnavailable: 2
  selector:
    matchLabels:
      app.kubernetes.io/name: hdfs
      app.kubernetes.io/instance: simple-hdfs
      app.kubernetes.io/component: datanode
      app.kubernetes.io/rolegroup: in-memory

Pros

  • easy to understand

  • easy to implement

  • works the same as all other config

Cons

  • Does not support the common use case of role level PDBs