|
| 1 | +// |
| 2 | +// DISCLAIMER |
| 3 | +// |
| 4 | +// Copyright 2018 ArangoDB GmbH, Cologne, Germany |
| 5 | +// |
| 6 | +// Licensed under the Apache License, Version 2.0 (the "License"); |
| 7 | +// you may not use this file except in compliance with the License. |
| 8 | +// You may obtain a copy of the License at |
| 9 | +// |
| 10 | +// http://www.apache.org/licenses/LICENSE-2.0 |
| 11 | +// |
| 12 | +// Unless required by applicable law or agreed to in writing, software |
| 13 | +// distributed under the License is distributed on an "AS IS" BASIS, |
| 14 | +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 15 | +// See the License for the specific language governing permissions and |
| 16 | +// limitations under the License. |
| 17 | +// |
| 18 | +// Copyright holder is ArangoDB GmbH, Cologne, Germany |
| 19 | +// |
| 20 | +// Author Ewout Prangsma |
| 21 | +// |
| 22 | + |
| 23 | +package resources |
| 24 | + |
| 25 | +import ( |
| 26 | + "context" |
| 27 | + "fmt" |
| 28 | + "time" |
| 29 | + |
| 30 | + "github.com/rs/zerolog" |
| 31 | + "k8s.io/api/core/v1" |
| 32 | + metav1 "k8s.io/apimachinery/pkg/apis/meta/v1" |
| 33 | + |
| 34 | + "github.com/arangodb/go-driver/agency" |
| 35 | + api "github.com/arangodb/kube-arangodb/pkg/apis/deployment/v1alpha" |
| 36 | + "github.com/arangodb/kube-arangodb/pkg/util/k8sutil" |
| 37 | +) |
| 38 | + |
| 39 | +// preparePodTermination checks if the given pod is allowed to terminate and if so, |
| 40 | +// prepares it for termination. |
| 41 | +// It returns nil if the pod is allowed to terminate yet, an error otherwise. |
| 42 | +func (r *Resources) preparePodTermination(ctx context.Context, log zerolog.Logger, p *v1.Pod, group api.ServerGroup, memberStatus api.MemberStatus, updateMember func(api.MemberStatus) error) error { |
| 43 | + var err error |
| 44 | + switch group { |
| 45 | + case api.ServerGroupAgents: |
| 46 | + err = r.prepareAgencyPodTermination(ctx, log, p, memberStatus) |
| 47 | + case api.ServerGroupDBServers: |
| 48 | + err = r.prepareDBServerPodTermination(ctx, log, p, memberStatus, updateMember) |
| 49 | + default: |
| 50 | + err = nil |
| 51 | + } |
| 52 | + return maskAny(err) |
| 53 | +} |
| 54 | + |
| 55 | +// prepareAgencyPodTermination checks if the given agency pod is allowed to terminate |
| 56 | +// and if so, prepares it for termination. |
| 57 | +// It returns nil if the pod is allowed to terminate, an error otherwise. |
| 58 | +func (r *Resources) prepareAgencyPodTermination(ctx context.Context, log zerolog.Logger, p *v1.Pod, memberStatus api.MemberStatus) error { |
| 59 | + // Inspect member phase |
| 60 | + if memberStatus.Phase.IsFailed() { |
| 61 | + log.Debug().Msg("Pod is already failed, safe to remove agency serving finalizer") |
| 62 | + return nil |
| 63 | + } |
| 64 | + // Inspect deployment deletion state |
| 65 | + apiObject := r.context.GetAPIObject() |
| 66 | + if apiObject.GetDeletionTimestamp() != nil { |
| 67 | + log.Debug().Msg("Entire deployment is being deleted, safe to remove agency serving finalizer") |
| 68 | + return nil |
| 69 | + } |
| 70 | + |
| 71 | + // Check node the pod is scheduled on |
| 72 | + agentDataWillBeGone := false |
| 73 | + if p.Spec.NodeName != "" { |
| 74 | + node, err := r.context.GetKubeCli().CoreV1().Nodes().Get(p.Spec.NodeName, metav1.GetOptions{}) |
| 75 | + if err != nil { |
| 76 | + log.Warn().Err(err).Msg("Failed to get node for member") |
| 77 | + return maskAny(err) |
| 78 | + } |
| 79 | + if node.Spec.Unschedulable { |
| 80 | + agentDataWillBeGone = true |
| 81 | + } |
| 82 | + } |
| 83 | + |
| 84 | + // Check PVC |
| 85 | + pvcs := r.context.GetKubeCli().CoreV1().PersistentVolumeClaims(apiObject.GetNamespace()) |
| 86 | + pvc, err := pvcs.Get(memberStatus.PersistentVolumeClaimName, metav1.GetOptions{}) |
| 87 | + if err != nil { |
| 88 | + log.Warn().Err(err).Msg("Failed to get PVC for member") |
| 89 | + return maskAny(err) |
| 90 | + } |
| 91 | + if k8sutil.IsPersistentVolumeClaimMarkedForDeletion(pvc) { |
| 92 | + agentDataWillBeGone = true |
| 93 | + } |
| 94 | + |
| 95 | + // Is this a simple pod restart? |
| 96 | + if !agentDataWillBeGone { |
| 97 | + log.Debug().Msg("Pod is just being restarted, safe to terminate agency pod") |
| 98 | + return nil |
| 99 | + } |
| 100 | + |
| 101 | + // Inspect agency state |
| 102 | + log.Debug().Msg("Agent data will be gone, so we will check agency serving status first") |
| 103 | + ctx = agency.WithAllowNoLeader(ctx) // The ID we're checking may be the leader, so ignore situations where all other agents are followers |
| 104 | + ctx, cancel := context.WithTimeout(ctx, time.Second*15) // Force a quick check |
| 105 | + defer cancel() |
| 106 | + agencyConns, err := r.context.GetAgencyClients(ctx, func(id string) bool { return id != memberStatus.ID }) |
| 107 | + if err != nil { |
| 108 | + log.Debug().Err(err).Msg("Failed to create member client") |
| 109 | + return maskAny(err) |
| 110 | + } |
| 111 | + if len(agencyConns) == 0 { |
| 112 | + log.Debug().Err(err).Msg("No more remaining agents, we cannot delete this one") |
| 113 | + return maskAny(fmt.Errorf("No more remaining agents")) |
| 114 | + } |
| 115 | + if err := agency.AreAgentsHealthy(ctx, agencyConns); err != nil { |
| 116 | + log.Debug().Err(err).Msg("Remaining agents are not healthy") |
| 117 | + return maskAny(err) |
| 118 | + } |
| 119 | + |
| 120 | + return nil |
| 121 | +} |
| 122 | + |
| 123 | +// prepareDBServerPodTermination checks if the given dbserver pod is allowed to terminate |
| 124 | +// and if so, prepares it for termination. |
| 125 | +// It returns nil if the pod is allowed to terminate, an error otherwise. |
| 126 | +func (r *Resources) prepareDBServerPodTermination(ctx context.Context, log zerolog.Logger, p *v1.Pod, memberStatus api.MemberStatus, updateMember func(api.MemberStatus) error) error { |
| 127 | + // Inspect member phase |
| 128 | + if memberStatus.Phase.IsFailed() { |
| 129 | + log.Debug().Msg("Pod is already failed, safe to remove dbserver pod") |
| 130 | + return nil |
| 131 | + } |
| 132 | + // Inspect deployment deletion state |
| 133 | + apiObject := r.context.GetAPIObject() |
| 134 | + if apiObject.GetDeletionTimestamp() != nil { |
| 135 | + log.Debug().Msg("Entire deployment is being deleted, safe to remove dbserver pod") |
| 136 | + return nil |
| 137 | + } |
| 138 | + |
| 139 | + // Check node the pod is scheduled on |
| 140 | + dbserverDataWillBeGone := false |
| 141 | + if p.Spec.NodeName != "" { |
| 142 | + node, err := r.context.GetKubeCli().CoreV1().Nodes().Get(p.Spec.NodeName, metav1.GetOptions{}) |
| 143 | + if err != nil { |
| 144 | + log.Warn().Err(err).Msg("Failed to get node for member") |
| 145 | + return maskAny(err) |
| 146 | + } |
| 147 | + if node.Spec.Unschedulable { |
| 148 | + dbserverDataWillBeGone = true |
| 149 | + } |
| 150 | + } |
| 151 | + |
| 152 | + // Check PVC |
| 153 | + pvcs := r.context.GetKubeCli().CoreV1().PersistentVolumeClaims(apiObject.GetNamespace()) |
| 154 | + pvc, err := pvcs.Get(memberStatus.PersistentVolumeClaimName, metav1.GetOptions{}) |
| 155 | + if err != nil { |
| 156 | + log.Warn().Err(err).Msg("Failed to get PVC for member") |
| 157 | + return maskAny(err) |
| 158 | + } |
| 159 | + if k8sutil.IsPersistentVolumeClaimMarkedForDeletion(pvc) { |
| 160 | + dbserverDataWillBeGone = true |
| 161 | + } |
| 162 | + |
| 163 | + // Is this a simple pod restart? |
| 164 | + if !dbserverDataWillBeGone { |
| 165 | + log.Debug().Msg("Pod is just being restarted, safe to remove dbserver pod") |
| 166 | + return nil |
| 167 | + } |
| 168 | + |
| 169 | + // Inspect cleaned out state |
| 170 | + log.Debug().Msg("DBServer data is being deleted, so we will cleanout the dbserver first") |
| 171 | + c, err := r.context.GetDatabaseClient(ctx) |
| 172 | + if err != nil { |
| 173 | + log.Debug().Err(err).Msg("Failed to create member client") |
| 174 | + return maskAny(err) |
| 175 | + } |
| 176 | + cluster, err := c.Cluster(ctx) |
| 177 | + if err != nil { |
| 178 | + log.Debug().Err(err).Msg("Failed to access cluster") |
| 179 | + return maskAny(err) |
| 180 | + } |
| 181 | + cleanedOut, err := cluster.IsCleanedOut(ctx, memberStatus.ID) |
| 182 | + if err != nil { |
| 183 | + return maskAny(err) |
| 184 | + } |
| 185 | + if cleanedOut { |
| 186 | + // Cleanout completed |
| 187 | + if memberStatus.Conditions.Update(api.ConditionTypeCleanedOut, true, "CleanedOut", "") { |
| 188 | + if err := updateMember(memberStatus); err != nil { |
| 189 | + return maskAny(err) |
| 190 | + } |
| 191 | + } |
| 192 | + // Trigger PVC removal |
| 193 | + if err := pvcs.Delete(memberStatus.PersistentVolumeClaimName, &metav1.DeleteOptions{}); err != nil { |
| 194 | + log.Warn().Err(err).Msg("Failed to delete PVC for member") |
| 195 | + return maskAny(err) |
| 196 | + } |
| 197 | + |
| 198 | + log.Debug().Msg("Server is cleaned out. Save to remove drain dbserver finalizer") |
| 199 | + return nil |
| 200 | + } |
| 201 | + // Not cleaned out yet, check member status |
| 202 | + if memberStatus.Conditions.IsTrue(api.ConditionTypeTerminated) { |
| 203 | + log.Warn().Msg("Member is already terminated before it could be cleaned out. Not good, but removing dbserver pod because we cannot do anything further") |
| 204 | + return nil |
| 205 | + } |
| 206 | + // Ensure the cleanout is triggered |
| 207 | + log.Debug().Msg("Server is not yet clean out. Triggering a clean out now") |
| 208 | + if err := cluster.CleanOutServer(ctx, memberStatus.ID); err != nil { |
| 209 | + log.Debug().Err(err).Msg("Failed to clean out server") |
| 210 | + return maskAny(err) |
| 211 | + } |
| 212 | + return maskAny(fmt.Errorf("Server is not yet cleaned out")) |
| 213 | +} |
0 commit comments