JakeHillion/storj
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

satellite/satellitedb: don't include DQd and exited nodes in DQStrayNodes

Browse Source 
Don't update DQ time of already DQd nodes. Don't DQ nodes who exited.

Change-Id: I4528a9ba9f8e278987165ad337a9b34dadb9788b
This commit is contained in:
Cameron Ayer 2021-02-10 08:58:05 -05:00
parent 3c49f0ff73
commit 549033f2e6
2 changed files with 43 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

44
satellite/overlay/db_test.go
View File

@ -21,22 +21,52 @@ import (
func TestDQNodesLastSeenBefore(t *testing.T) { func TestDQNodesLastSeenBefore(t *testing.T) {
testplanet.Run(t, testplanet.Config{ testplanet.Run(t, testplanet.Config{
SatelliteCount: 1, StorageNodeCount: 1, SatelliteCount: 1, StorageNodeCount: 2,
}, func(t *testing.T, ctx *testcontext.Context, planet *testplanet.Planet) { }, func(t *testing.T, ctx *testcontext.Context, planet *testplanet.Planet) {
node := planet.StorageNodes[0] planet.Satellites[0].Overlay.DQStrayNodes.Loop.Pause()
node.Contact.Chore.Pause(ctx)
node1 := planet.StorageNodes[0]
node2 := planet.StorageNodes[1]
node1.Contact.Chore.Pause(ctx)
node2.Contact.Chore.Pause(ctx)
cache := planet.Satellites[0].Overlay.DB cache := planet.Satellites[0].Overlay.DB
info, err := cache.Get(ctx, node.ID()) // initial check that node1 is not disqualified
n1Info, err := cache.Get(ctx, node1.ID())
require.NoError(t, err) require.NoError(t, err)
require.Nil(t, info.Disqualified) require.Nil(t, n1Info.Disqualified)
// gracefully exit node2
req := &overlay.ExitStatusRequest{
NodeID: node2.ID(),
ExitInitiatedAt: time.Now(),
ExitLoopCompletedAt: time.Now(),
ExitFinishedAt: time.Now(),
}
dossier, err := cache.UpdateExitStatus(ctx, req)
require.NoError(t, err)
require.NotNil(t, dossier.ExitStatus.ExitFinishedAt)
// check that node1 is disqualified and node2 is not
require.NoError(t, cache.DQNodesLastSeenBefore(ctx, time.Now())) require.NoError(t, cache.DQNodesLastSeenBefore(ctx, time.Now()))
info, err = cache.Get(ctx, node.ID()) n1Info, err = cache.Get(ctx, node1.ID())
require.NoError(t, err) require.NoError(t, err)
require.NotNil(t, info.Disqualified) require.NotNil(t, n1Info.Disqualified)
n1DQTime := n1Info.Disqualified
n2Info, err := cache.Get(ctx, node2.ID())
require.NoError(t, err)
require.Nil(t, n2Info.Disqualified)
// check that node1 DQ time is not overwritten
require.NoError(t, cache.DQNodesLastSeenBefore(ctx, time.Now()))
n1Info, err = cache.Get(ctx, node1.ID())
require.NoError(t, err)
require.Equal(t, n1DQTime, n1Info.Disqualified)
}) })
} }

8
satellite/satellitedb/overlaycache.go
View File

@ -1553,10 +1553,14 @@ func (cache *overlaycache) populateUpdateFields(dbNode *dbx.Node, updateReq *ove
return updateFields return updateFields
} }
// DQNodesLastSeenBefore disqualifies all nodes where last_contact_success < cutoff. // DQNodesLastSeenBefore disqualifies all nodes where last_contact_success < cutoff except those already disqualified
// or gracefully exited.
func (cache *overlaycache) DQNodesLastSeenBefore(ctx context.Context, cutoff time.Time) (err error) { func (cache *overlaycache) DQNodesLastSeenBefore(ctx context.Context, cutoff time.Time) (err error) {
defer mon.Task()(&ctx)(&err) defer mon.Task()(&ctx)(&err)
q := `UPDATE nodes SET disqualified = current_timestamp WHERE last_contact_success < $1;` q := `UPDATE nodes SET disqualified = current_timestamp
WHERE last_contact_success < $1
AND disqualified is NULL
AND exit_finished_at is NULL;`
_, err = cache.db.ExecContext(ctx, q, cutoff) _, err = cache.db.ExecContext(ctx, q, cutoff)
return err return err
} }