Customers have noted individual machines may frequently fail backups, especially during initial full backup on boarding. This article discusses some ideas that may impact success
Especially during a first initial full backup, customers may note the backup will disconnect of fail frequently. This article discusses some impacts and things that can be done to limit these disconnections.
The primary reason an initial backup will fail is the loss of an active network connection. There are various reasons this will occur. This list is not intended to be conclusive:
- Upload speed has permitted less than 10MB of data transfer in a 30 minute period.
- WiFi is unstable, impeded, resets, or is completely disrupted for more than 60 seconds.
- The connected network changes resulting in a new IP issued or a change in network security policy applied
- A firewall or other security device is monitoring upload traffic and has imposed a throttle or stopped the connection due to defined QoS or other security policy limitations
- The system has entered sleep or hibernate states during a backup
- The machine is powered off
- System resources are too low resulting in difficulty reading data, or windows/application stability issues
- Antivirus software interference in processes
To ensure your connection for initial full backup is reliable, ensure the following are reviewed:
- Disable hibernate and sleep functions, or set them to longer terms that will ensure the backup is not interrupted until complete. Once initial full backups are done, daily backups will typically take only a few minutes to a few hours depending on amount of data changed and speed of connection.
- Do not completely log out of windows, simply lock the session for security until initial backup is complete.
- Do not power off the machine or close it’s lid (laptops) during initial backup.
- Ensure the PC is not starved for resources. When backups are not active, the system should have sufficient RAM and resources to maintain little to no pagefile use. Disk or CPU should not sustain 100% utilization for more than brief periods of time.
- Check firewall settings to ensure the PC is not subject to maximum bandwidth or data cap uploads per security policy. Trying the backup from a different location that is not subject to the same firewall may reveal certain networks have such caps in place.
- During the initial full backup, attempt if possible to leave the system at rest so it is not changing network segments during the backup time. It may be necessary to leave the system plugged in and online for many days for the initial backup to complete.
- On busy or older wifi systems, try backing up using physical cabling/hardwire instead.
- Set antivirus scan policies to exempt from scan all Unitrends processes located in C:\PCBP. AV products that perform scan on read may significantly exacerbate system performance during backup and cause disruptions even to local enterprise class backup systems.
- Reduce the data set of the system being protected to only critical paths and folders that contain user data, including ensuring you exclude windows and application directories, multimedia folders, and any other large amount of data that would not typically be restored in a disaster.
Additional backup retries do not have to re-send all data to the Unitrends EndPoint Cloud, but, must have sufficient bandwidth to rescan data that was prior sent and encounter sufficient new data within 30 minutes to meet thresholds. On some slow connections, there may be insufficient bandwidth to complete these operations and begin data transmission within the 30-minute period, especially on machines with more data.
Though Unitrends cannot make a specific recommendation on bandwidth requirements to successfully complete backups, as this will vary wildly by machine, network, and nature of data, connections with sustained or filtered uploads that fall below 10Mbit/sec may suffer significantly more issues than others. Systems that are unable to reliably upload all data in less than 7 days over their connection may also wish to temporarily find faster connections, or, limit the scope of data protected.
Once an initial full is complete and successful, significantly less daily bandwidth is required, sufficient only to transmit all new created or modified blocks to the cloud in less than 24 hours.