ORtera Atlas 1.1.12 Release Notes

Introduction

These release notes describe changes in this release compared to previous releases, as well as known issues. For installation and operation instructions, please see the README file and the User Guide, available online at http://ortera.com as well as in the Compass GUI.

Known Issues

• When tens of thousands of files are being accessed during a monitoring session, processing the result may take very long time, as the filename reconstruction and mapping algorithm is ineffecient at handling such a large number of filenames.
• When tens of thousands of files are being accessed during a monitoring session, a Java 1.4.2 internal error may cause the JVM to exit. Upgrading to Java 5 solves this problem.
• The Managed Host software must run as root.
• Compass can get completely stuck when it can't communicate with the server, e.g., because the server is shut down or because of network problems. As a work-around, kill the Compass process and start a new one.
• On very busy systems, the user may experience a delay between scheduling a monitoring session and receiving confirmation that the monitoring session has been scheduled. One work-around, to continue scheduling more monitoring sessions, is to open a second Compass window. Occasionally the confirmation never appears although the monitoring session has been scheduled. To continue scheduling kill and restart Compass, or refresh that Compass window.
• Under very rare circumstances, for instance, when discovery analysis data is being saved at exactly the same time as the monitoring or heuristics analysis data for the same application or filesystem is being saved, one of the database operations (and hence the analysis phase it is part of) may fail. As a work-around, rerun the analysis job.

Changes Since ORtera Atlas 1.1.11

• Support for files greater than 2GB has been added. In the process of monitoring filesystems each active file is checked to match its device to the device of the monitored filesystem. The stat() system call was failing for files larger than 2 GB causing the file to be excluded from analysis. The stat64 () system call is now used to ensure all files are included when they are active on a monitored filesystem.

Changes Since ORtera Atlas 1.1.10

• A few potential stability-related problems have been fixed

Changes Since ORtera Atlas 1.1.9

• Under rare circumstances, the atlasanalyze command (used to process the monitoring data) could fail with a core dump. This problem has been corrected and the code has been optimized for faster processing.
• The processing of monitor data in atlasagentd has been optimized for faster processing.
• ORtera performs an automatic analysis of I/O intensive processes when the Managed Host process is started for the first time, see the User Guide for details. Under certain circumstances, the job counters displayed in Compass were not reset when this job finishes, resulting in one job incorrectly shown as staying in the Discovery phase. This problem has been corrected.

Changes Since ORtera Atlas 1.1.8

• Links in /dev/rdsk that point to non-existing targets are now ignored instead of causing "discovery" to fail.
• The Summary Report is now generated as HTML, and includes heuristics information. It opens in the default web browser, and can be saved and printed using the standard browser features.
• Additional status information for the "monitoring" phase has been added. A new "Processing monitoring data" event is logged to the Event Log when the actual monitoring is completed, and the Status Bar now shows counters for both "monitoring" and "processing."
• A job progress indicator has been added to the Status Bar. It is active as long as at least one job on a Managed Host is in an active state.
• The $ATLAS_HOME/bin/atlas_start.sh script now verifies that the database is working before starting ORtera Atlas.

Changes Since ORtera Atlas 1.1.7

• The bandwidth figures in the Compass Performance -> Bandwidth and Response Time tab have been refined for certain types of workload, and the instantaneous interarrival time in the Compass Workload -> Time Utilization tab calculation has been refined. Note! The ORtera Atlas server part is unchanged and does not need to be upgraded for this version.

Changes Since ORtera Atlas 1.1.6

• Analyzed nodes for which there was no I/O activity during the monitoring session are now shown with gray icons in the Compass tree, to make it easier to find the relevant processes with data in the tree. Note! The ORtera Atlas server part is unchanged and does not need to be upgraded for this version.

Changes Since ORtera Atlas 1.1.5

• The monitoring phase could fail for short monitoring sessions on a very busy server. This issue has been fixed.
• The error messages in the Event Log and the server log file now contain text explaining the failure in addition to the return code.
• The startup time for Managed Hosts has been reduced.

Changes Since ORtera Atlas 1.1.4

• Temporary instrumentation data files were left in the $ATLAS_VAR/wfns-root/raw directory when running the Managed Host software on Solaris 8. These files are now removed correctly for all Solaris versions, unless they are explicitly requested to be kept with the atlasagentd "-kwf" option.

Changes Since ORtera Atlas 1.1.2

• The atlasanalyze command has been extended to produce information about mmap() activity, and to correctly handle the case where all I/O activity consists of mmap() calls. See the atlasanalyze man page for details.
• A new heuristic has been added to indicate that lack of regular I/O requests at the filesystem level of the storage infrastructure may be because memory mapping is used by the application being monitored. The ORtera atlasanalyze command line tool can be used to analyze the amount of latency caused by memory mapping in your configuration.
• Compass has been improved to handle errors during CSV Import correctly, and to label all Resource Set nodes in the tree with the resource type name.

Changes Since ORtera Atlas 1.1.1

• A serious bug in the ESFS, affecting operation on Solaris 8 only, has been fixed. The du(1M) command would incur an endless loop if applied to the ESFS mount point. This condition is detected and avoided in later version of Solaris, but not on Solaris 8. The fix allows the du(1M) command to work properly on the ESFS mount point even on Solaris 8. ESFS is a pseudo file system, so the output of the du(1M) command is not of interest, but automated scripts checking disk usage in general may check the ESFS mount point because it is there.
• Monitoring duration time has been limited to 30 minutes in Compass. Monitoring sessions exceeding 30 minutes rarely have any incremental information value beyond shorter durations, and data for such long sessions can take a very long time to process on some systems. Monitoring sessions of any duration can still be performed using the command-line interface.
• The root filesystem is no longer available for analysis in the Filesystems tab, because it typically does not hold data of interest for ORtera analysis, and monitoring the root filesystem slows processing of monitoring results.

Changes Since ORtera Atlas 1.1

• The port number for atlasagentd reported at start-up was incorrect on Solaris 8. This has been corrected.
• atlasagentd could core dump on shutdown in some circumstances on Solaris 8. This has been corrected.

Changes Since ORtera Atlas 1.0.2

• The Heuristics analysis now detects the following additional conditions:
  • Unused Physical Devices
  • Over-Allocation of Physical Resources
  See the corresponding sections in the User Guide for details.
• The Heuristics analysis now provides different recommendations depending on resource type (files, logical or physical devices) for the following conditions:
  • High Peak Load
  • Large I/O Size Variation
  • Unbalanced Data
  • Unbalanced I/O Operations
  • Unbalanced Response Times
  See the corresponding sections in the User Guide for details.

Changes Since ORtera Atlas 1.0.1

• Under rare circumstances, I/O for certain files was not recorded. This has been fixed.
• I/O request records for the same file could in some cases be split between records registered under the real filename and an encoded filename. This has been fixed.
• Some users reported that heuristics failed occasionally on high-end servers (more than 30 CPUs) for analysis sessions of a longer duration (e.g., 10 minutes or more), reported as a NullPointerException in the Compass Event Log. We have changed the software to prevent this from happening.
• Refreshing the Discovered Resources node in Compass didn't get the most up-to-date data. This has been fixed.
• The Time column values in the I/O Per Resource tab are now formatted consistently to always have 3 decimal places.
• The I/O Per Resources and the Resources tabs now show the total number of resources.
• The script usage messages shown during installation of the server part, and when the user provides incorrect server type options to the atlas_start.sh script, incorrectly stated that the "ah" option requires root privileges. This has been fixed. Now they state that it is the "mh" option that requires root privileges.

Changes Since ORtera Atlas 1.0

• A bug in the Compass Import CSV Files feature, causing the Properties tabs for the imported Application and Application Partition nodes to be blank and the imported node to appear directly under the root note instead of under the Imported Results node, has been fixed. Note! The ORtera Atlas server part is unchanged and does not need to be upgraded for this fix.
• The ORtera Atlas x86 server code now supports both 32 bit and 64 bit processors. This change only affects the x86 server part.

Changes Since ORtera Atlas 0.38 (Beta 5)

• Partial support for NFS has been added, making it possible to monitor I/O activity for the NFS mount point. If the NFS server is also an ORtera Atlas Managed Host, the I/O activity through all storage resource layers for the exported filesystem can be monitored on the NFS server, and the I/O for the mount point on the remote server can then be manually compared to the I/O at the NFS server.
• The installer for the server software (pkgadd) now creates Bash and C shell profiles for the ortera account, setting all ORtera Atlas environment variables and creating useful aliases.
• The atlas_startup.sh script now requires the server type (mh, ah or both) to be specified explicitly, to avoid starting the server as the default type by mistake.
• The Heuristics analysis condition and recommendation texts have been clarified and edited for readability. Links to sections in the User Guide, describing each heuristics in detail, have also been added to the heuristics list in the Heuristics tab in Compass.
• Help buttons have been added in the Processes, Filesystems and Analyze tabs, launching the User Guide with the appropriate section selected.
• The User Guide has been updated to cover all features, with new sections describing all heuristics in detail. All screenshots have also been replaced with up-to-date versions, and the Installation section has been updated to reflect the current distribution file structure and naming conventions.
• Clicking on a link to an external web site in the User Guide now launches the native web browser, so that the site can be viewed with a full-featured web browser.
• Compass on Mac OS X can now launch the native web browser for accessing web sites through the Help menu and the User Guide. This feature was previously only available for Windows, Solaris and Linux.
• The Properties tab for imported Application and Filesystem nodes now shows the name of the imported file.
• Various minor bug fixes in both Compass and the server code.

Changes Since ORtera Atlas 0.37 (Beta 4)

• The number of Managed Hosts and the total number of scheduled and running jobs in different phase are now shown in the Compass Status Bar
• When Atlas is started on a Managed Host for the first time, it now automatically performs an initial analysis of the filesystems used by the most I/O intensive processes, in addition to the analysis it previously did for the most I/O intensive processes.
• The Compass Event Log window now pops up automatically when you connect to an Aggregator Host, and its initial location is maintained between Compass sessions.
• Compass now provides helpful hints about what the different nodes represent and what you can do at each level in the Compass tree.
• The Compass Help menu now contains direct links to various pages on the ORtera web site.
• System error messages in the Compass Event Log now contain a reference to the analysis job that failed.
• An issue with the Heuristics analysis that caused it to fail (reported as a NullPointerException error message in the Event Log), in some environments under certain circumstances, has been corrected.
• Various stability-related bug fixes

Changes Since ORtera Atlas 0.36 (Beta 3)

• Extended the Export to XML File feature to also work on individual Application Partition nodes. See the User Guide for details.
• The default path for the dynamic ORtera Atlas data is now /var/opt/ortera/atlas instead of /var/ortera/atlas
• Fixed various stability issues in the ORtera backend
• Fixed problems with lost data at the filesystem level for processes with gaps in the file descriptor list (typical for Oracle processes when using filesystems for the database).
• Fixed problems with attribution of I/O to the wrong filesystem when using a subdirectory in one filesystem as the mount point for another.
• Fixed a problem where KB values in the Resource Set Properties tab and the I/O Size tab in Compass ended up as 0 instead of as a fractional number for small values.
• Added a clarification label for Load Level in the Load Level tab in Compass
• Fixed refreshing of Application Partition nodes in Compass so that the Last Monitored field gets updated correctly.

Changes Since ORtera Atlas 0.35 (Beta 2)

• The result of each analysis job is now saved as a separate Application Partition, with the name prepended with a timestamp. Along with this change, we also made the discovery phase mandatory (can't be deselected) because it's the phase that creates the default partition. The Analyze tab was moved from the Application Partition node to the Application/Filesystem node, since that is actually what we analyze today.
• Individual Application/Filesystem and Application Partition nodes can now be deleted; just select one or more nodes and choose Delete Nodes in the Edit menu.
• Added View Summary Report for Application Partition nodes, see the User Guide for details.
• Added a Rename menu item in the Edit menu and a right-mouse click pop-up menu.
• The Resource Set name is now displayed instead of the ID in the Heuristics tab Owner column (now renamed to Category), and this column is now only present in the Heuristics tab for Application Partition nodes
• Added I/O Per Resource sub-tab under the Workload tab, showing Ops/Data/Time per used resource.
• Added sorting for all tables except the Event Log table, with a sort-order indication in the column header. It's now also possible to sort on multiple columns: click on the main column and then Ctrl-click on the additional columns.
• Compass errors are now reported in pop-up in addition to the status bar message.
• Fixed the inconsistency in the Workload and Performance tabs layout of the statistics with regards to the order of Min, Max, Avg, etc. (reported by Dominic Kay, Sun Microsystems). The order is now Min, Max, central tendencies, variations in all cases.
• The Properties tabs for Aggregator and Managed Host nodes now show the OS, Java and Atlas versions
• PID and Host info added to the Application/Filesystem Properties tabs
• Added scrollbar to the Properties tab
• Added a Properties tab for Imported Results
• Changed the size unit from bytes to KB (with 3 decimals) in all tabs.
• The ORtera Atlas, Java and OS versions are now included in Java log files on both the server and the client.
• The df command output for the ESFS pseudo filesystem now shows Used as 0% instead of 100% (reported by Dominic Kay, Sun Microsystems).
• Updated and expanded the User Guide content to accurately describe the functionality available in the current version, and addressed the issues reported by Dominic Kay, Sun Microsystems.
• Various minor adjustments and bug fixes

Changes Since ORtera Atlas 0.33.14 (Beta 1)

• Added support for Solaris 10
• Replaced the numbers in the Compass Workload tab columns with node names
• Modified the default URLs for the Aggregator/Managed Hosts to use IP addresses instead of host names, to allow for access to an Aggregator Host from Compass in another domain by default.
• Modified all Compass Properties tabs to contain customized data for the selected node type, for instance, an analysis result summary for Resource Set nodes, alias only for node types that can be renamed, and more detailed type information for Resource nodes.
• Added an Imported Results top-level node to group all imported XML/CSV files in Compass, and changed the Compass behavior to not remove the imported results when connecting to a new Aggregator Host.
• Added a Remove From Tree menu item in the Compass Edit menu, for selective removal of imported results.
• Added more threads for processing the raw instrumentation data resulting in a performance improvement when monitoring many processes/filesystems at the same time.
• Various stability-related bug fixes