Network Backups: Common Pitfalls

First of all make sure the network drive is NTFS formatted. If it is not, there will be critical limitations, depending on the file system used. NTFS allows paths to reach over 32,000 letters, whereas Linux file systems don’t. Linux file systems have different file size limits and file name limits (not just full path length limits). Access control lists, aka file / folder permissions, are also stored differently on NTFS than other systems that may not provide the same permission structure for files.

DNS bugs and issues: lots of NAS devices have issues with network name resolution. To work around that issue, use static IP addresses throughout and use the IP address in the UNC paths: \\10.0.0.23\myshare

Backup user permissions. BackupChain uses the local system account to access files and folders. Neither admins nor the local system user have access to everything. Especially in a domain environment, the best way to avoid issues is to create a dedicated admin user “BackupChain” with a fixed, never expiring password and switch the BackupChain Service to that user. In addition, you need to give that new domain admin user explicit access to everything.

When you have many network shares on the same network server that give access to different users per share, we recommend using the same strategy. Create a new dedicated BackupChain user and give access to all shares on that server. The reason is that Windows does not allow several users to authenticate to the same network server coming from one identical server. For example, you can’t authenticate users A and B logged on to server G where they work, can’t both access network shares on server H using their respective logins. The way around it, they would both have to use the same user to connect to H because Microsoft does not allow multiple logins to the same network server.

Instability under high loads: Some NAS devices can be quite instable. Under high loads connections may break with bizarre errors. Sometimes the issue is not the NAS box itself. Check cabling and switches for defects by replacing one at a time.

Transfer speeds: Ethernet won because it was cheap, not great. When you have several network servers transferring files over the network, the total bandwidth throughput dramatically decreases due to packet collisions. This design isn’t very useful when it comes to optimizing throughput; but since almost everyone uses Ethernet you need to work around it by scheduling backups to run at separate intervals and thereby avoid overlaps.

Link delays: even 10Gbps Ethernet is not as fast as a locally attached drive via SATA for example. In order for data to travel to the other server it needs to be repackaged many times in to packets, etc. Also wire lengths and switch buffering and caching all delay packets measurably. When you back up a lot of small files the overhead becomes an issue. The “ping-pong” of requests and response per file is a real issue when you have 10 million files to back up. Therefore, one needs to plan ahead and expect much longer backup times when backups are to be sent over the wire.

Standby and power off issues: Some consumer grade NAS servers are peculiar in their own way. When backups pause for a while they may turn off and not return properly when activity resumes, or they may break existing links before they do so. It’s best to turn off power saving mode on those kinds of devices.