Hi there,
I have a deployment using Azure blob for the cloud object store @ 12.0. I am seeing ArcGIS Server backups fail once they get over a certain size, or if I attempt to include object store caches in the backups. I get this message in the webgisdr logs:
{
"name": "ArcGIS Server",
"messages": "Cannot access the backup file for ArcGIS Server in webgisdr-backup/12-0-0/backup/20260505-091107-backup\\server\\cd67fdf9jcwpzzk6.",
"URL": "https://xyzxyzxy/hosting",
"status": "failure"
}
Latest job status of async backup operation on server https://xyzxyzx/hosting: {"jobid":"j65e79912-4aaa-46d3-8b8d-0717760d431b","stepInfo":["Export operation started.","Export operation failed."],"status":"EXECUTING"}.
"{"jobid":"j65e79912-4aaa-46d3-8b8d-0717760d431b","status":"COMPLETED","stepInfo":["Export operation started.","Export operation failed."],"operationResponse":{"status":"error","messages":["Export operation failed. Failed to write the file '20260505-091124.agssite' to this location 'webgisdr-backup/12-0-0/backup/20260505-091107-backup/server/cd67fdf9jcwpzzk6'"]}}"
looking at a manual export of that site from the admin API I get:
Failed to write the file '20260505-095011.agssite' to this location 'backup-20260504' Failed to upload backup-20260504/20260505-095011.agssite to object store.. Caused by: Timeout on blocking read for 300000000000 NANOSECONDS
Export operation failed. Failed to write the file '20260505-095011.agssite' to this location 'backup-20260504'
which is an exception in:
Caused by: Timeout on blocking read for 300000000000 NANOSECONDS at com.esri.arcgis.objectstore.clients.azure.AzureBlobStoreClient.putObject(AzureBlobStoreClient.java:986) at com.esri.arcgis.objectstore.clients.azure.AzureBlobStoreClient.putObject(AzureBlobStoreClient.java:915) at com.esri.arcgis.discovery.admin.backuprestore.BackupRestoreUtil.writeSiteFileToBackupStore(BackupRestoreUtil.java:888)
This worked at 11.5, there was a change from staging everything locally first before uploading to writing directly to blob storage at 12.0. We are using private endpoints to access the storage and should have a clear path to it.
I have seen two bugs with 12.0 but don't know if this is related to what I'm seeing:
If anyone has hit this have you got a cause / solution? I'm going to staging locally and uploading to blob in the meantime.
Thanks!
UPDATE: I'm trying to write the backup to a local drive / share now and have updated the webgisdr properties to suit. If I run with no Azure config in the webgisdr.properties file I get this:
{"backupLocation":"","messages":"The value of the property BACKUP_BLOB_ACCOUNT_NAME cannot be empty.","id":"cbcfbc64699144a89586a1bb1383ba80","operation":"backup","results":[],"status":"failure","elapsedTime":"00hr:00min:02sec"}
and if I put in the location of the objstore container I get this in the server logs and the export fails:
Could not connect to the Backup Object Store specified. Check if the specified parameters are valid. JSONObject["objectStore"] not found.
this is with INCLUDE_OBJECT_STORE_CACHES = true
Another update - if I set:
# Specify whether to include object store caches.
INCLUDE_OBJECT_STORE_CACHES = false
then I can get a .agssite file locally.
Hi,
did you find a solution to backup with the object store?
I am currently having the same issue. It seems like the tool assumes the object store is a cloud store. In the server admin, the cloud store is also listed under cloudStores: server/admin/data/items/cloudStores/AGSDataStore_objectstore_oz_suwxku9
Hi, unfortunately no answer yet though I have logged an issue with Esri. What version are you at? At 12.0 backing up the object store seemed busted - I tested under 12.1 and it was backing up again ok. There is also an issue if you register a cloud store (say for cache locations) at 12.0 where it thinks it should be an object store - I haven't tested that at 12.1.
Hi, I think that was the problem in my case. I'm on 11.5, the object store is on-prem with the dedicated object store role for datastore. I had an azure cloud store attached to the arcgis server, but for caches, not as an object store. I think it assumed it's an object store as well and tries to back it up. When I removed the azure store (it was for testing purposes only), it worked.
Also note there is an issue backing up to blob storage with big webgisdr files where this times out. This I think is fixed in 12.1 and I think there is a patch coming for 12.0.