Chapter 6. Troubleshooting a multisite Ceph Object Gateway

This chapter contains information on how to fix the most common errors related to multi-site Ceph Object Gateways configuration and operational conditions.

Note

When the bucket sync status command reports bucket is behind on shards even if the data is consistent across multi-site, performing additional writes to the bucket, synchronizes the sync status reports and displays the message bucket is caught up with source.

6.1. Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • A running Ceph Object Gateway.

6.2. Error code definitions for the Ceph Object Gateway

The Ceph Object Gateway logs contain error and warning messages to assist in troubleshooting conditions in your environment. Some common ones are listed below with suggested resolutions.

Common error messages

data_sync: ERROR: a sync operation returned error
This is the high-level data sync process complaining that a lower-level bucket sync process returned an error. This message is redundant; the bucket sync error appears above it in the log.
data sync: ERROR: failed to sync object: BUCKET_NAME:_OBJECT_NAME_
Either the process failed to fetch the required object over HTTP from a remote gateway or the process failed to write that object to RADOS and it will be tried again.
data sync: ERROR: failure in sync, backing out (sync_status=2)
A low level message reflecting one of the above conditions, specifically that the data was deleted before it could sync and thus showing a -2 ENOENT status.
data sync: ERROR: failure in sync, backing out (sync_status=-5)
A low level message reflecting one of the above conditions, specifically that we failed to write that object to RADOS and thus showing a -5 EIO.
ERROR: failed to fetch remote data log info: ret=11
This is the EAGAIN generic error code from libcurl reflecting an error condition from another gateway. It will try again by default.
meta sync: ERROR: failed to read mdlog info with (2) No such file or directory
The shard of the mdlog was never created so there is nothing to sync.

Syncing error messages

failed to sync object
Either the process failed to fetch this object over HTTP from a remote gateway or it failed to write that object to RADOS and it will be tried again.
failed to sync bucket instance: (11) Resource temporarily unavailable
A connection issue between primary and secondary zones.
failed to sync bucket instance: (125) Operation canceled
A racing condition exists between writes to the same RADOS object.

Additional Resources

6.3. Syncing a multisite Ceph Object Gateway

A multisite sync reads the change log from other zones. To get a high-level view of the sync progress from the metadata and the data logs, you can use the following command:

Example

[ceph: root@host01 /]# radosgw-admin sync status

This command lists which log shards, if any, which are behind their source zone.

Note

Sometimes you might observe recovering shards when running the radosgw-admin sync status command. For data sync, there are 128 shards of replication logs that are each processed independently. If any of the actions triggered by these replication log events result in any error from the network, storage, or elsewhere, those errors get tracked so the operation can retry again later. While a given shard has errors that need a retry, radosgw-admin sync status command reports that shard as recovering. This recovery happens automatically, so the operator does not need to intervene to resolve them.

If the results of the sync status you have run above reports log shards are behind, run the following command substituting the shard-id for X.

Syntax

radosgw-admin data sync status --shard-id=X --source-zone=ZONE_NAME

Example

[ceph: root@host01 /]# radosgw-admin data sync status --shard-id=27 --source-zone=us-east
{
  "shard_id": 27,
  "marker": {
         "status": "incremental-sync",
         "marker": "1_1534494893.816775_131867195.1",
         "next_step_marker": "",
         "total_entries": 1,
         "pos": 0,
         "timestamp": "0.000000"
   },
   "pending_buckets": [],
   "recovering_buckets": [
         "pro-registry:4ed07bb2-a80b-4c69-aa15-fdc17ae6f5f2.314303.1:26"
   ]
}

The output lists which buckets are next to sync and which buckets, if any, are going to be retried due to previous errors.

Inspect the status of individual buckets with the following command, substituting the bucket id for X.

Syntax

radosgw-admin bucket sync status --bucket=X.

Replace X with the ID number of the bucket.

The result shows which bucket index log shards are behind their source zone.

A common error in sync is EBUSY, which means the sync is already in progress, often on another gateway. Read errors written to the sync error log, which can be read with the following command: 

radosgw-admin sync error list

The syncing process will try again until it is successful. Errors can still occur that can require intervention.

6.3.1. Performance counters for multi-site Ceph Object Gateway data sync

The following performance counters are available for multi-site configurations of the Ceph Object Gateway to measure data sync:

  • poll_latency measures the latency of requests for remote replication logs.
  • poll_errors measures the number of errors from poll requests.
  • avgtime is the average latency of poll requests.
  • fetch_bytes measures the number of objects and bytes fetched by data sync.
  • fetch_errors measures the number of errors from fetch requests.
  • fetch_not_modified measures the number of times sync tried to fetch an object, but found that it already has the latest copy.

Use the ceph --admin-daemon command to view the current metric data for the performance counters:

Syntax

ceph --admin-daemon /var/run/ceph/ceph-client.rgw.RGW_ID.asok perf dump data-sync-from-ZONE_NAME

Example

[ceph: root@host01 /]# ceph --admin-daemon /var/run/ceph/ceph-client.rgw.host02-rgw0.103.94309060818504.asok perf dump data-sync-from-us-west

{
    "data-sync-from-us-west": {
        "fetch bytes": {
            "avgcount": 54,
            "sum": 54526039885
        },
        "fetch not modified": 7,
        "fetch errors": 0,
        "poll latency": {
            "avgcount": 41,
            "sum": 2.533653367,
            "avgtime": 0.061796423
        },
        "poll errors": 0
    }
}

Note

You must run the ceph --admin-daemon command from the node running the daemon.

Additional Resources

  • See the Ceph performance counters chapter in the Red Hat Ceph Storage Administration Guide for more information about performance counters.

6.4. Synchronizing data in a multi-site Ceph Object Gateway configuration

In a multi-site Ceph Object Gateway configuration of a storage cluster, failover and failback causes data synchronization to stop. The radosgw-admin sync status command reports that the data sync is behind for an extended period of time.

You can run the radosgw-admin data sync init command to synchronize data between the sites and then restart the Ceph Object Gateway. This command does not touch any actual object data and initiates data sync for a specified source zone. It causes the zone to restart a full sync from the source zone.

Important

Contact Red Hat support before running the data sync init command.

If you are going for a full restart of sync, and if there is a lot of data that needs to be synced on the source zone, then the bandwidth consumption is high and then you have to plan accordingly.

Note

If a user accidentally deletes a bucket on the secondary site, you can use the metadata sync init command on the site to synchronize data.

Prerequisites

  • A running Red Hat Ceph Storage cluster.
  • Ceph Object Gateway configured at two sites at least.

Procedure

  1. Check the sync status between the sites:

    Example

    [ceph: host04 /]# radosgw-admin sync status
          realm d713eec8-6ec4-4f71-9eaf-379be18e551b (india)
      zonegroup ccf9e0b2-df95-4e0a-8933-3b17b64c52b7 (shared)
           zone 04daab24-5bbd-4c17-9cf5-b1981fd7ff79 (primary)
   current time 2022-09-15T06:53:52Z
zonegroup features enabled: resharding
  metadata sync no sync (zone is master)
      data sync source: 596319d2-4ffe-4977-ace1-8dd1790db9fb (secondary)
                        syncing
                        full sync: 0/128 shards
                        incremental sync: 128/128 shards
                        data is caught up with source

  2. Synchronize data from the secondary zone:

    Example

    [ceph: root@host04 /]# radosgw-admin data sync init --source-zone primary

  3. Restart all the Ceph Object Gateway daemons at the site:

    Example

    [ceph: root@host04 /]# ceph orch restart rgw.myrgw