Failed to backup to ceph(s3) using xtrabackup #8391
Comments
|
we met the same issue that the compressed file cannot be upload into OSS. Could anyone provide support? Urgent for me ,thanks! |
We have solved the problem. We have changed
vttablet \
$TOPOLOGY_FLAGS \
-log_dir $VTDATAROOT/tmp \
-log_queries_to_file $VTDATAROOT/tmp/$tablet_logfile \
-tablet-path $alias \
-tablet_hostname "$tablet_hostname" \
-init_keyspace $keyspace \
-init_shard $shard \
-init_tablet_type $tablet_type \
-health_check_interval 5s \
-enable_semi_sync \
-enable_replication_reporter \
-backup_storage_implementation s3 \
-s3_backup_aws_endpoint "http://north1.minio.local" \
-s3_backup_force_path_style \
-s3_backup_storage_root "/root/data" \
-s3_backup_storage_bucket "commerce" \
-backup_engine_implementation xtrabackup \
-xtrabackup_root_path "/usr/bin" \
-xtrabackup_user root \
-xtrabackup_stream_mode xbstream \
-restore_from_backup \
-port $port \
-grpc_port $grpc_port \
-service_map 'grpc-queryservice,grpc-tabletmanager,grpc-updatestream' \
-pid_file $VTDATAROOT/$tablet_dir/vttablet.pid \
-vtctld_addr http://$hostname:$vtctld_web_port/ \
> $VTDATAROOT/$tablet_dir/vttablet.out 2>&1 &
|
|
I can confirm the same behavior -- it seems to potentially be an issue with the minio version the client uses. |
|
@grsmith-projects does that mean the built-in backup works? Have you tried that? |
|
@askdba built-in does not work The backup succeeds, the errors occur when the files start being transferred -- only 1 file succeeds and the rest fail due to the connection being signalled to exit. Also, we rebuilt the system w/ various versions of minio lib (6.x and 7.x) and had the same issue. My gut feeling right now is that some default setting for miniolib might be conflicting with the way newer ceph does multipart uploads -- we are currently investigating tuning on the ceph side. Minio client itself works Minio client from a container in the same pod works.. but vtbackup does not. (All of this is speculation at this point) |
|
from ceph.go: Could it be that your server does not support v2 clients? |
deepthi
added
Component: Backup and Restore
and removed
Component: Cluster management
labels
|
I can reproduce the issue with
I tried with |
Overview
Hi~, I try the local demo , everything was fine at first, but the error happend when I backed an instance up.
Only one
MANIFESTfile is stored on OSS.backup.xbstream.gznot found.And error information
So I need some help, THANK YOU!
Version Infomation
Reproduction Steps
Steps to reproduce this issue, example:
/etc/ceph_conf.json){ "accessKey": "minioadmin", "secretKey": "minioadmin", "endPoint": "north1.minio.local:7000", "useSSL": false }My Question
MANIFESTupload successfully means OSS work well. I don't understand why tar upload failed. Maybe I ignore some flags or version Incompatibility.Please give me some advice.
If need more logs and information, I can provide it.
THANK YOU!
The text was updated successfully, but these errors were encountered: