Misc. Local Environment Tasks

This page contains information about software and tasks specific to our local Digital UNIX environment. It is not meant to be an exhaustive treatment of each topic, just a quick introduction to some common tasks for each topic.

EMC Symmetrix

Sybase

OS Backups and the Tape library

Server Management software

This Web Server


EMC Symmetrix

The EMC Symmetrix is a multi-host disk storage array. It can be connected to multiple hosts of a variety of operating systems. At Reynolds, the array is connected to systems via fast wide differential scsi controllers. It is currently (6/98) connected to: oh02ux03, oh02ux20, oh02ux21, oh02ux22, oh02ux24, oh02nt13, and oh02nt16. From the operating system point of view, the Symmetrix looks like any other locally attached SCSI disks.

EMC Maintenance issues

The Symmetrix performs extensive self maintenance and monitoring. It uses idle IO bandwidth to perform pro-active surface scans of the disks and CRC checks of the systems internal cache, and error diagnostics of the systems other components. The intention of which is to spot hardware errors in the early stages when they are still retryable rather than waiting for them to turn into hard errors. In the event that an error is found, the Symmetrix will dial back to the EMC support center an notify the center of the error. EMC support personnel will then dial into the system and evaluate the situation, dispatching a field tech if needed. Due to the fault tolerance built into the system, Reynolds' first indication of a problem would most likely be a call from our local EMC support office.

EMC Symmetrix software

Because the Symmetrix uses standard SCSI interfaces to connect to our hosts, no special drivers are needed on the host side to access disks on the Symmetrix. However, there is software needed for software features such as performance monitoring, Business Continuance Volumes (aka Time Finder) or Symmetrix Remote Data mirroring Facility (SRDF). This software can manage the Symmetrix from any system that is connected to it. As of this writing (6/98), Reynolds is licensed to use the UNIX version of this software, and can load it on any of the UNIX systems connected to the Symmetrix. It is currently loaded on oh02ux22 and oh02ux24 under the /usr/symmapps directory. It is important to have the software on more than one system so that the BCVs can be manipulated even if one of the systems is down

Configuring new disks residing on the Symmetrix

Because the Symmetrix does not require any special drivers, configuring new disk drives on a host does not require any special processes. Disks can be configured into the system using the "/dev/MAKEDEVS" script in Digital UNIX. EMC does however provide a script to make this process easier. It is called "/dev/MAKE_EMC_LUNS", and it will create the new disk devices for all the partitions of the disk, set the desired device ownerships, and install an initial disk label.

In order to allow greater numbers of SCSI disks to be connected to a given SCSI controller, EMC utilizes the "LUN" value that is part of normal SCSI disk addressing, but is not commonly used. This means that a disk may have a target of 0, and a LUN of 0, giving it a device name of /dev/rz0, and you may also have a disk with a target of 0 and a LUN of 1, giving it a device name of /dev/rzb0.


Sybase

Configuring new raw disks for Sybase

Configuring new raw disk partitions for Sybase is an pretty simple process. There are just a couple things to remember:

Sybase Backups

Sybase backups are performed by the Sybase backupserver process rather than by using standard operating system tools. This allows an internally consistent backup to be performed even while the database is on-line and active. The backups are run using SQL commands. The actual processing and is performed by the "backupserver" process. Because of this, a "ps" command would not show any activity going to the process that runs the backup script.

Configuration issues for Sybase

Two important configuration issues for Sybase involve the amount of memory it uses and the number of "engines" that the server will use. Both of these are controlled from the Sybase point of view by the start-up configuration file that lives in the "$SYBASE/install" directory. The memory parameter controls the total amount of memory that sybase will use. Most of which is allocated in 1 big shared memory segment at start-up. This means that the "shm_max" parameter in the operating system as well as the "max-per-proc-data-size" and "max-per-proc-address-space" kernel parameter should be configured to large enough values to contain this data.

The other important parameter, max engines, controls how many server processes Sybase will start up. This is an method of throttling the amount of CPU that a sybase server will consume. The default value is to only start up a single server process, which means that on a multi-cpu machine, Sybase will only consume a fraction of the total CPU time available on the system. Depending on the other processes running on the box, you may want to configure Sybase to use as many engines as there are CPUs on the system. On modern SMP systems, Sybase should peacefully coexist with the operating system and other processes even at 100% cpu utilization.


OS Backups and the Tape library

NSR Software

We have Networker Save and Restore software licensed that are not currently in use. The original system architecture involved using oh02ux22 as a central backup server utilizing it's tape library unit with 4 DLT drives, and a 48 slot tape handling robot.

Controlling the tape library

The tape library attached to oh02ux22 comes with a command line and a GUI interface for manipulating the robot. The command line interface is "/usr/bin/robot" and the GUI interface is "/usr/bin/X11/xrobot". There is a man page for both of these commands on oh02ux22.


Server Management software


This Web Server

There are two components to the documentation web server. Most of the content of the server is served up by the freely available Apache web server. The other server that also runs on this box is a copy of the DynaWeb SGML to HTML server. This server is provided with Sybooks and is used to provide the online web browser interface to Sybooks. These two servers are seperate daemons on the same system with their own configuration and data files.

Apache Server

The Apache server is a copy of the freely available web server available from the Apache project web site, www.apache.org. This web server is well supported via Internet resources and is the server that runs most of the web servers on the Internet.

Its directory Structure follows that of a typical web server:
Installation root dir: /home/httpd
Configuration files: " /conf
Root document dir: " /htdocs
cgi-bin dir: " /cgi-bin
log files " /logs

This server was compiled here from version 1.2.4 source code. The source code is archived in oh02ux23:/home/admin/archive.

The server has a start up and shutdown script in /sbin/init.d, "httpd". It runs under a user id of "nobody" and listens for requests on the standard http port of 80.

DynaWeb / Sybooks

In order to server up Sybooks via a web server, you need to run a specific web server software package by DynaBooks. This server provides search facillities and converts the Sybooks from their native format (SGML) to HTML on the fly. This server runs on oh02ux23 along side the Apache server. In order to run both servers on the same system, it is neccessary to run the DynaWeb server on a different port than the Apache server. It listens for requests on port 8080.

It's start up script is /sbin/init.d/sybooks.

This is the directory structure for the Sybooks / DynaWeb installation:
Sybooks root dir: /usr/sybase/sybooks
Sybooks config file: " /.ebtrc
DynaWeb root dir: " /server
DynaWeb config dir " /data/config

The main administrative task for maintaining the Sybooks installation is the installation of new Sybooks Collections. This procedure is not intuitively obvious. The Sybooks installation program that we have installed has a bug in it that makes it unreliable when adding new collections. The DynaWeb configuration program then has problems due to this previous bug. To add new book collections, I have been doing it manually.