|HP-UX System Administrator's Guide: Routine Management Tasks: HP-UX 11i Version 3 > Chapter 2 Booting and Shutdown
Hopefully, most of your system shutdowns will be of this type. With a normal shutdown, you have time to prepare the system and its users so that the system can be restarted and work can continue with no loss of data, and as little disruption as possible.
In order to maximize system performance, recently used data from disk files is kept and updated in memory. Periodically (by default, every 30 seconds), a program called sync is run to make sure the file systems on disk are kept up to date in the event of an unplanned shutdown (the on-disk file systems are synchronized with the memory-based changes). But, if it’s been 29 seconds since the last run of sync, there are probably memory based changes that are not yet reflected on disk. If the system crashes now, this can cause inconsistencies in file system structures on disk (which, although not usually the case, can cause corrupt files or loss of data).
Also, users of both your system and other systems in the network that depend on your system for some resource will be affected. It is always best to notify them in advance of any planned shutdown so that they can plan for the shutdown and minimize the impact to their work.
Many HP-UX systems can be equipped with uninterruptible power supplies (UPSs) to allow you to maintain power to your systems for a short while following the failure of your computer’s primary power source. If the power failure is brief, systems equipped with UPSs will not be affected by the power failure at all. If the power failure appears as though it will last for a long time, you can use the buffer period provided by an uninterruptible power supply to perform a normal shutdown. See “Normal (Planned) Shutdown”.
Computers equipped with HP PowerTrust uninterruptible power supplies can also be monitored by a special daemon called upsmond, which, when running, always resides in memory (is not swappable). upsmondcommunicates with the power supplies, and when power has been off for longer than a pre-configured time period, upsmond will perform a clean shutdown of your system automatically.
Not all HP-UX systems are equipped with uninterruptible power supplies. If yours is not, an unclean shutdown is the likely result of a power failure. No memory dump will be performed, and it is possible that buffers of recent disk changes still reside in memory, and have not been written to disk by the sync program. See “Unclean Shutdowns” for details.
When a power failure occurs, it is good practice to turn off the power switches to your computer and its peripherals. This will reduce the chances of a power surge harming your equipment when the power comes back on. After the power is restored, follow normal boot procedures. See “A Standard Boot ”.
When a system is properly shut down, all memory-based file system changes are written to disk and the file systems on disk are marked as being clean. However, if an improper shutdown (for example, a power failure) occurs, the memory-based information might not be written to disk and therefore certain file systems will not have their “clean” flag set (because, in fact, they might have structural problems as a result of the memory-based information not being written to disk).
When this happens, a special activity occurs during the boot process. The file system consistency checker (fsck), when checking for clean flags on all file systems represented in the file /etc/fstab, will detect that file systems exist that do not have clean flags set. For these file systems, fsck will perform a check/repair operation to locate and fix any problems that resulted from the improper shutdown. In nearly all cases, fsck can find and fix all of the structural problems and the file system can then be marked clean.
On rare occasions, the file system corruption is beyond what fsck can automatically correct. In these cases fsck will terminate with an error message indicating that you need to use it in an interactive mode to fix the more serious problems. In these cases data loss is likely. Before using fsck in interactive mode, try to back up any critical files by moving them to another file system or backing them up to tape, if a backup copy of them does not already exist.
Although rare, sometimes systems can shut themselves down unexpectedly in an event known as a system crash or system panic. For a detailed description of what to do if this happens, and an explanation of what takes place following a system crash, see “Configuring Dump Devices”.
A special operating mode, called single-user mode, is available on HP-UX systems. While your system is in single-user mode only the console is active, and a lot of the subsystems for HP-UX are not running. This mode is usually used for system maintenance. There are two ways to put your system into single-user mode:
In today’s world of networked computers, people who are not direct users of your system can still be affected by its absence from the network (when it has been shut down). If your system is serving one or more of the following functions, you need to at least consider the impact to users of other systems when you plan to take your system down; and, if possible, you should try to let them know in advance that they will be affected, so that they can prepare for the event.
If your system is a mail server, it receives e-mail on behalf of its users, and is often the computer handling the outgoing e-mail for them too. When your system is down, incoming mail is usually held by other computers in the network for delivery when your system is back on line. If your computer will be down for an extended period of time, it is possible that others sending e-mail to your computer’s users will have their e-mail returned as being undeliverable.
If your computer is a network name server (for example, a DNS name server), it is responsible for translating computer alias names into IP addresses for its own users and those who have configured their systems to use your computer as their name server. Usually systems are configured to use multiple sources for name switch information so if your system is down, they can use an alternate name server, a local hosts file, or directly use IP addresses to access remote machines until your system is back on line.
If your computer is serving as a network gateway computer: that is, it has several network interface cards in it, and is a member of multiple networks (subnets), your computer’s absence on the network can have a huge impact on network operations. An example of this is the computer called flserver in the Sample Network. While such a computer is down, computers on one of the subnets are unable to communicate with computers on other subnets, unless other gateway computers exist that can handle the traffic.
If your computer is an NFS file server, other computers in the network may have mounted one or more of your computer’s file systems to be a part of their own directory trees. When your system goes down, attempts to access the files or directories of your system by users on the other systems will result in those accesses hanging if the file systems have been hard mounted, or returning an error if they have been soft mounted. After your system reboots the client systems may operate normally or return a stale file handle error. If a stale file handle error occurs, you can unmount then remount the file system, the other systems will likely require a reboot once your system is back on line before those systems will again be able to access your computer’s file systems.
The best course of action is to alert the administrators of systems who have NFS-mounted file systems from your computer to unmount the NFS-mounted file systems before you shut down your system! By doing this, they will simply need to re-mount the NFS file systems from your computer when your computer is back on line. No reboot of the other systems will be required.
Provided that NFS clients are not also acting as NFS servers for other computers (computer B in the preceding note is acting as both NFS client and server), it is safe to shut them down without affecting the NFS server. It will simply be necessary to remount the file system from the NFS server when the NFS client has rebooted. This is probably done automatically during the boot-up process.
As described earlier, there are times when a normal, planned shutdown is appropriate. But as server downtime becomes less desired and accepted, on-line addition and replace functionality can help you to avoid shutting down a server in many cases.
HP-UX’s On-line Addition and Replacement of PCI Cards (OLA/R) features enable you to replace a faulty interface card or add a new interface card to a running system, without impacting the system’s users.