[arvados] updated: 2.1.0-2867-g4d417144e

[git repository hosting]

Summary of changes:
 lib/cloud/loopback/loopback.go | 9 +++++++++
 1 file changed, 9 insertions(+)

       via  4d417144e48b0f4c11efd744954a2f9e115fb9ab (commit)
      from  10440ac12d6771ab80469adf551d2cac8d3461e6 (commit)

Those revisions listed above that are new to this repository have
not appeared on any other notification email; so we list those
revisions in full, below.


commit 4d417144e48b0f4c11efd744954a2f9e115fb9ab
Author: Tom Clegg <tom at curii.com>
Date:   Tue Sep 6 12:11:12 2022 -0400

    17344: Reset "broken node" flag on loopback instance create/reset.
    
    Arvados-DCO-1.1-Signed-off-by: Tom Clegg <tom at curii.com>

diff --git a/lib/cloud/loopback/loopback.go b/lib/cloud/loopback/loopback.go
index 6ad4f876d..9b71c9f16 100644
--- a/lib/cloud/loopback/loopback.go
+++ b/lib/cloud/loopback/loopback.go
@@ -11,6 +11,7 @@ import (
 	"encoding/json"
 	"errors"
 	"io"
+	"os"
 	"os/exec"
 	"os/user"
 	"strings"
@@ -58,6 +59,14 @@ func (is *instanceSet) Create(it arvados.InstanceType, _ cloud.ImageID, tags clo
 	if len(is.instances) > 0 {
 		return nil, errQuota
 	}
+	// A crunch-run process running in a previous instance may
+	// have marked the node as broken. In the loopback scenario a
+	// destroy+create cycle doesn't fix whatever was broken -- but
+	// nothing else will either, so the best we can do is remove
+	// the "broken" flag and try again.
+	if err := os.Remove("/var/lock/crunch-run-broken"); err != nil && !errors.Is(err, os.ErrNotExist) {
+		return nil, err
+	}
 	u, err := user.Current()
 	if err != nil {
 		return nil, err

-----------------------------------------------------------------------


hooks/post-receive
--