18071: Remove obsolete comment.

[arvados.git] / services / keep-balance / balance.go

diff --git a/services/keep-balance/balance.go b/services/keep-balance/balance.go
index fa01d512bc77b20d3ad871f21b14c5c88611ab0f..9f581751d938baf8d0f8fdc726bfae92e1f6877d 100644 (file)
--- a/services/keep-balance/balance.go
+++ b/services/keep-balance/balance.go
@@ -2,12 +2,13 @@
 //
 // SPDX-License-Identifier: AGPL-3.0
 
-package main
+package keepbalance
 
 import (
        "bytes"
        "context"
        "crypto/md5"
+       "errors"
        "fmt"
        "io"
        "io/ioutil"
@@ -22,7 +23,9 @@ import (
        "syscall"
        "time"
 
+       "git.arvados.org/arvados.git/lib/controller/dblock"
        "git.arvados.org/arvados.git/sdk/go/arvados"
+       "git.arvados.org/arvados.git/sdk/go/ctxlog"
        "git.arvados.org/arvados.git/sdk/go/keepclient"
        "github.com/jmoiron/sqlx"
        "github.com/sirupsen/logrus"
@@ -66,16 +69,19 @@ type Balancer struct {
 // subsequent balance operation.
 //
 // Run should only be called once on a given Balancer object.
-//
-// Typical usage:
-//
-//   runOptions, err = (&Balancer{}).Run(config, runOptions)
-func (bal *Balancer) Run(client *arvados.Client, cluster *arvados.Cluster, runOptions RunOptions) (nextRunOptions RunOptions, err error) {
+func (bal *Balancer) Run(ctx context.Context, client *arvados.Client, cluster *arvados.Cluster, runOptions RunOptions) (nextRunOptions RunOptions, err error) {
        nextRunOptions = runOptions
 
+       ctxlog.FromContext(ctx).Info("acquiring active lock")
+       if !dblock.KeepBalanceActive.Lock(ctx, func(context.Context) (*sqlx.DB, error) { return bal.DB, nil }) {
+               // context canceled
+               return
+       }
+       defer dblock.KeepBalanceActive.Unlock()
+
        defer bal.time("sweep", "wall clock time to run one full sweep")()
 
-       ctx, cancel := context.WithDeadline(context.Background(), time.Now().Add(cluster.Collections.BalanceTimeout.Duration()))
+       ctx, cancel := context.WithDeadline(ctx, time.Now().Add(cluster.Collections.BalanceTimeout.Duration()))
        defer cancel()
 
        var lbFile *os.File
@@ -266,6 +272,29 @@ func (bal *Balancer) CheckSanityEarly(c *arvados.Client) error {
                }
        }
 
+       mountProblem := false
+       type deviceMount struct {
+               srv *KeepService
+               mnt *KeepMount
+       }
+       deviceMounted := map[string]deviceMount{} // DeviceID -> mount
+       for _, srv := range bal.KeepServices {
+               for _, mnt := range srv.mounts {
+                       if first, dup := deviceMounted[mnt.DeviceID]; dup && first.mnt.UUID != mnt.UUID && mnt.DeviceID != "" {
+                               bal.logf("config error: device %s is mounted with multiple volume UUIDs: %s on %s, and %s on %s",
+                                       mnt.DeviceID,
+                                       first.mnt.UUID, first.srv,
+                                       mnt.UUID, srv)
+                               mountProblem = true
+                               continue
+                       }
+                       deviceMounted[mnt.DeviceID] = deviceMount{srv, mnt}
+               }
+       }
+       if mountProblem {
+               return errors.New("cannot continue with config errors (see above)")
+       }
+
        var checkPage arvados.CollectionList
        if err = c.RequestAndDecode(&checkPage, "GET", "arvados/v1/collections", nil, arvados.ResourceListParams{
                Limit:              new(int),