NOTICE: This version of the NSF Unidata web site (archive.unidata.ucar.edu) is no longer being updated.
Current content can be found at unidata.ucar.edu.
To learn about what's going on, see About the Archive Site.
Dear LDM user, Version 6.8.0 of the LDM is now available from the LDM homepage at <http://www.unidata.ucar.edu/software/ldm/> or from <ftp://ftp.unidata.ucar.edu/pub/ldm/ldm-6.8.0.tar.gz> This version features the following: 1) The way the top-level LDM server vets new client connections is improved; 2) More information is printed by the "ldmadmin config" command; 3) The accuracy of the system clock can be checked by ldmadmin(1); and 4) LDM performance metrics can now be collected and plotted by the ldmadmin(1) script if the requisite utilities are installed. VETTING NEW CLIENT CONNECTIONS: In this version, the way the top-level LDM server vets new client connections is improved. Before, the LDM server would get the hostname of the client system and then check both the IP address and hostname of the client system against the ALLOW and ACCEPT entries in the LDM configuration-file before fork(1)ing a child LDM process to handle the connection. Now, the LDM server immediately and unconditionally fork(2)s a child LDM process to handle the connection and the child process uses a two-phase procedure to vet the client system. First the IP address of the client is checked against the ALLOW and ACCEPT entries. If that succeeds, then the client connection is allowed; otherwise, the hostname of the client system is obtained and that is then checked against the entries. These changes were made to prevent a client whose hostname cannot be found from suspending the LDM server until the hostname lookup times-out. As a consequence of these changes, there is now an upper limit on the number of active client connections to an LDM server. The default is 256. This value can be changed via the "$max_clients" variable in the ldmadmin(1) configuration-file (etc/ldmadmin-pl.conf). CHECKING THE SYSTEM CLOCK: As you may know, an accurate system clock is necessary for correct operation by the LDM: an inaccurate clock can cause a delay in data reception or even a loss of data. By default, this new feature causes almost every ldmadmin(1) command to also check the accuracy of the system clock. This new feature is also highly customizable: everything from being disabled to causing it to *not* start the LDM if the clock is off by a user-settable threshold. The default is to check the system clock but to allow the LDM to start. To customize this feature, pay close attention to the instructions on web-page <http://www.unidata.ucar.edu/software/ldm/ldm-current/basics/vet-ldmadmin.ht ml>. Do a case-insensitive search for the word "time". ACCUMULATING AND PLOTTING METRICS: This feature deals with metrics for the LDM system such as the number of LDM connections, the age of the oldest data-product in the product-queue, CPU load average and modes, etc. This feature allows the metrics to be periodically accumulated in a file and to be displayed as time-series plots (great for figuring-out if your product-queue is large enough). You will need to have the top(1), uptime(1), netstat(1), and vmstat(1) utilities installed in order to collect metrics, and you'll need to have the gnuplot(1) utility installed in order to display them. To use this feature, pay close attention to the web-pages <http://www.unidata.ucar.edu/software/ldm/ldm-current/basics/platform.html>, <http://www.unidata.ucar.edu/software/ldm/ldm-current/basics/vet-ldmadmin.ht ml>, <http://www.unidata.ucar.edu/software/ldm/ldm-current/basics/configuring.htm l>, and <http://www.unidata.ucar.edu/software/ldm/ldm-current/basics/monitoring.html >. Do a case-insensitive search for the word "metric". For an extensive history of changes to the LDM, see the file CHANGE_LOG in the top-level source-directory. As always, we are happy to install the LDM system for you if you grant us root access to your system. Regards, Steve Emmerson LDM Developer
ldm-users
archives: