KP_CLUSTERTOOLS, UTILITIES, VMS and Cluster system management tools CONTIG.COM Creates or extends a system file in as contiguous a manner as possible. COMBINE_FEEDBACK.COM Logically combines feedback data from multiple files. DECNET_LATENCY.COM Measure round-trip network latency over DECnet FIXQUE.COM Creates DCL code to re-create queues, jobs, forms, and characteristics, based on output from $SHOW QUEUE. LANTRAF.COM Displays relative amount of SCS traffic on LAN adapters NODE$TRACK.COM Track cluster membership and notify someone of changes REXMT.COM Shows packet retransmit counters to detect LAN problems SHAD.COM Shows shadowset members which are in a merge or copy SHAD_TRACK.COM Shows shadowset merge/copy progress over time SIFT_LAVC.COM Creates time-sorted, cluster-wide list of messages from SYS$EXAMPLES:LAVC$FAILURE_ANALYSIS.MAR program. SHADOW$CONFIG.COM Gathers shadow-set state info into DCL symbols or logical names, for use by other procedures, like these: SHADOW$DISPLAY.COM Uses ASCII art to display color-coded diagrams showing shadowset state in multi-site clusters. SHAD$DISP.COM Version with abbreviated, more-compact output, and optional support for disaster-tolerant clusters. SHADOW$TRACK.COM Track shadowset membership and notify someone of changes SHOW_PREF_PATH.COM Show preferred LAN path being used to each cluster node More-detailed descriptions: CONTIG.COM Creates or extends a system file in as contiguous a manner as possible. System files can only have one file header (no extension headers) because they are accessed by primitive code. In a Contiguous Best Try file extension, VMS will find the three largest extents on the disk, but then gives up and proceeds to allocates space arbitrarily for the rest of the file. This means the file header for a system file may fill up prematurely, sometimes before the desired file size is reached, and a backup/restore of the entire disk is then required to repack the file before it can be extended to the desired size. This DCL procedure finds the largest free extent and enlarges the file by only that amount each cycle, repeating this until the desired size is reached, thus minimizing the degree of fragmentation for the resulting file, and minimizing the number of extents used up in the file header. COMBINE_FEEDBACK.COM Logically combines feedback data from multiple files. Some systems (especially government systems) are not allowed to run continuously for the 24 hours or more that AUTOGEN wants for valid feedback data. These systems may run 8 hours per day, then be shut down and have their disks removed and locked in a safe overnight. This DCL procedure logically combines the data from multiple AUTOGEN feedback files, so that if feedback data is saved at the end of each uptime session, these periods of data can effectively be combined together and used as input to AUTOGEN as if they were contiguous. The procedure basically works by adding the values from cumulative counters together, finding the highest value for peak values, and using the most recent data for any data items which represent "current" data, and adding the elapsed times together. DECNET_LATENCY.COM Sends 100 loopback packets using NCP, and measures the average round-trip network latency. Since it uses the NCP> LOOP NODE command, it needs DECnet Phase IV. FIXQUE.COM Creates DCL code to re-create queues, jobs, forms, and characteristics, based on output listing from $SHOW QUEUE. Was developed to solve queue file corruption; still useful to help in moving queues and jobs between systems or setting up one cluster to have the same queue environment as another. Is also supplied in a "squeezed" version (FIXQUE_SQZ.COM) for faster execution. LANTRAF.COM Display the LAN adapters enabled for SCS traffic and the relative amounts of traffic SCS has sent to each NODE$TRACK.COM Track cluster membership and notify someone by e-mail when things change, like when a node leaves the cluster or joins the cluster. REXMT.COM Examines PEDRIVER's retransmitted-packet counters under SDA, to measure the severity of LAN problems in an LAVc environment. ReXmt counter shows re-transmitted packets (packet was lost or its Ack was lost). ReRcv counter shows duplicate packets received (Ack was lost on its way back to sender). Both indicate packet-loss problems in the LAN. SHAD.COM Displays shadowset members which are in a shadow merge or shadow copy state, if any. SHAD_TRACK.COM Periodically displays shadowset members which are in a shadow merge or shadow copy state, and degree of completion, to allow tracking copy or merge progress over time. SIFT_LAVC.COM Sifts through OPERATOR.LOG files on all nodes in a cluster and creates a time-sorted, consolidated list of all the %LAVC console messages from SYS$EXAMPLES:LAVC$FAILURE_ANALYSIS.MAR in the format of: dd-mmm-yyyy hh:mm:ss.cc nodexx %LAVC-s-ident, text This allows one to more easily determine the sequence of events and timing in a failure of a cluster interconnect. SHADOW$CONFIG.COM Determine the existence, membership, and status of all shadowsets on the system, and their member devices. Stores results in either DCL symbols or in logical names. The following information is returned by the procedure: SHADOW$CONFIG_set_count Number of shadowsets SHADOW$CONFIG_SET_n_LABEL Volume label for each shadowset SHADOW$CONFIG_SET_n_DEVICE_NAME DSAn: device name for shadowset SHADOW$CONFIG_SET_n_MEMBER_COUNT shadowset size (# of members) SHADOW$CONFIG_SET_n_MEMBER_m_DEVICE_NAME Member device names SHADOW$CONFIG_SET_n_MEMBER_m_STATE Member state: "ShadowMember", "FullCopying", or "MergeCopying" There is also a timestamp logical name defined if the results are placed into logical names: SHADOW$CONFIG_TIMESTAMP Time of data-gathering SHADOW$DISPLAY.COM Display, in semi-graphical format, all the host-based volume shadowing shadow sets, highlighting any anomalies (such as lack of redundancy or full copies in progress) with the judicious use of color. SHAD$DISPLAY.COM Abbreviated version of SHADOW$DISPLAY.COM, taking less on-screen real estate. This version optionally supports a multi-site disaster-tolerant cluster configuration (it does assume that disk allocation classes are $1$ at one site and $2$ at the other site so it can tell which site a given disk is located at). Shadowsets without cross-site shadowing protection are then flagged as unprotected, even if they have two or more members, if all are within a single site. SHADOW$TRACK.COM Track shadowset membership and notify someone by e-mail when a member is added or lost, when a new shadowset is created, or when a shadowset goes away. SHOW_PREF_PATH.COM Show the preferred paths being used by PEDRIVER to each remote node in the cluster. ------------------------------------------------------------------------------ Author: Keith Parris parris@encompasserve.org or keithparris@yahoo.com Feel free to direct questions or comments to me, or send me improved versions. Look for latest updated versions or fixes at: http://encompasserve.org/~parris/ or http://www.geocities.com/keithparris/