PowerFlex: GET_INFO - Support Bundle Collection Utility

get_info.sh [OPTIONS]

DESCRIPTION

get_info.sh is a diagnostic utility that collects debug information from a PowerFlex (formerly ScaleIO) host and archives it into a compressed bundle for analysis by support personnel.

The utility gathers data from multiple sources, including:

• PowerFlex component logs, configuration, and trace files
• MDM/SCLI query outputs and internal debug dump
• PowerFlex component internal diagnostics
• Operating system configuration, logs, and runtime state
• Hardware inventory (storage controllers, network devices, NVMe, NVDIMM, and so forth.)
• Core dumps (existing and optionally generated on demand)
• Diagnostic data collector (diag_coll) statistics

The resulting bundle is a single compressed archive (tar/gz by default) that can be transferred to PowerFlex support for further analysis.

Only one instance of get_info.sh can run on a host at a given time. If there is not enough free space for its output, it refuses to run (unless space checking is explicitly skipped).

OPTIONS

General Options

• -a, --all

  Collect all data. This is equivalent to specifying --mdm-repository, --collect-cores, --max-cores=2, --valgrind-cores, and --analyse-diag-coll.

• -A, --analyse-diag-coll

  Analyze diagnostic data collector (diag coll) data

• -b[COMPONENTS], --collect-cores[=COMPONENTS]

  Collect existing core dumps for the space-separated list of user-land COMPONENTS. Default (when COMPONENTS is omitted): All user-land components
   

Example

-b'mdm sds'

--collect-cores='mdm sds'

• -d OUT_DIR, --output-dir=OUT_DIR

  Store the resulting bundle under directory OUT_DIR. Default: <WORK_DIR>/scaleio-getinfo (see --work-dir).

• -f, --skip-mdm-login

  Skip the query for PowerFlex MDM login credentials. Useful when the user has already logged in manually.

• -h, --help

  Show the help message and exit. When combined with --tech, also display technician options.

• -J, --xz

  Use tar/xz format for the collected bundle instead of the default tar/gz. Ignored if the system's tar(1) does not support --use-compress-program or xz(1) is not found.

• -k NUM, --max-cores=NUM

  Collect up to NUM core files from each component. Default: All core files, implies --collect-cores

• -l, --light

  Generate a light bundle. Only the latest generation of numbered log files is collected, and component executables or libraries are not included when collecting cores. Use of this option reduces supportability and its use is discouraged.

• -m NUM, --max-traces=NUM

  Collect up to NUM PowerFlex trace files from each component. Default: All files

• -N, --skip-space-check

  Skip free disk space verification before data collection.

• -P PATH, --collect-path=PATH

  Collect the additional path PATH. Only absolute paths are accepted. Accepts wildcards; wildcards should be quoted. This option can be specified multiple times to collect multiple paths.

• -q, --quiet, --silent

  Suppress messages on standard output.

• -r, --mdm-repository

  Collect MDM repository files.

• -s, --skip-sdbg

  Skip collection of SDBG (diagnostic debugger) output.

• -S, --pause-core-generation

  Pause core generation of PowerFlex components during data collection. Original configuration is restored after collection completes.

• -w WORK_DIR, --work-dir=WORK_DIR

  Use directory WORK_DIR for temporary files. Default: /tmp.

• -x FILE, --output-file=FILE

  Store the collected bundle as file named FILE. The appropriate file name suffix (.tgz, .zip, and so forth) is added automatically. If FILE is - (dash), write the bundle to standard output (implies --quiet). When a bundle is written to standard output, no bundle file is created on disk. Default: getInfoDump.

• -z, --zip

  Use zip format for the collected bundle instead of the default tar/gz. Ignored if zip(1) is not found on the system.

• --mdm-port=PORT

  Connect to the MDM using port PORT for SCLI commands. Default: scli default behavior.

• --overwrite-output-file

  Overwrite the output file if it already exists. When an output file or directory is explicitly specified (using -x or -d), the default behavior is to refuse to overwrite; this option overrides that.

• --tech

  Include technician options in the help message output.

MDM Login Options

The following options are passed to the SCLI --login command. Their behavior and default values are governed by SCLI.

• -n, --use-nonsecure-communication

  Connect to the MDM in nonsecure mode.

• -p PASSWORD, --password=PASSWORD

  Use PASSWORD for PowerFlex MDM login. Default: SCLI default behavior

• -u USERNAME, --username=USERNAME

  Use USERNAME for PowerFlex MDM login. Default: SCLI default behavior

• --ldap-authentication

  Log in to PowerFlex MDM using LDAP-based authentication.

• --management-system-ip=ADDRESS

  Connect to SSO/M&O at ADDRESS for PowerFlex login. Default: SCLI default behavior

• --p12-password=PASSWORD

  Encrypt the PowerFlex login PKCS#12 file using PASSWORD. Default: SCLI default behavior

• --p12-path=FILE

  Store the PowerFlex login PKCS#12 file as FILE. Default: SCLI default behavior

Technician Options

The following options are intended for use by support technicians and are shown in the help message only when --tech is specified.

• -c[COMPONENTS], --generate-cores[=COMPONENTS]

Generate core files (using gcore(1)) for the running processes of the space-separated list of user-land COMPONENTS. Default: all user-land components, implies --collect-executables, and requires gdb and gcore.

Examples:

-c'mdm sds'
--generate-cores='mdm sds'

• -C CORE_FILE, --reference-core-file=CORE_FILE

  Collect product logs and cores relative to the last modification time (mtime) of CORE_FILE, instead of the execution start time. Implies --collect-cores

• -E REF_TIME, --event-time=REF_TIME

  Collect product logs and cores relative to REF_TIME, instead of the execution start time. Accepts any format understood by date(1). Implies --collect-cores.

• -g[COMPONENTS], --valgrind-cores[=COMPONENTS]

Collect valgrind core dumps for the specified user-land COMPONENTS. Default: all user-land components. Implies --collect-executables.

Examples:

-g'mdm sds'
--valgrind-cores='mdm sds'

• -t MIN, --minutes-before-event=MIN

  Collect product logs and cores generated up to MIN minutes before the reference time. Default: 15

• -T MIN, --minutes-after-event=MIN

  Collect product logs and cores generated up to MIN minutes after the reference time. Default: 5

• -X[COMPONENTS], --collect-executables[=COMPONENTS]

  Collect component executables and their shared libraries for the specified user-land COMPONENTS. Default: All user-land components

Examples:

-X'mdm sds'
--collect-executables='mdm sds'

• --keep-work-dir

  Retain the generated temporary work directory after bundle creation (cleaned up automatically).

BUNDLE STRUCTURE

The output bundle is a single compressed archive.

• The bundle top-level directory is the hostname of the collected system.
• General host command outputs go into a server/ subdirectory.
  File name is <command> + <arguments> + suffix (.txt by default). Spaces replaced with _, nonalphanumeric characters stripped
  Example: server/ip_-s_addr.txt – output of ip -s addr
• Product command outputs go into the component’s subdirectory. mdm/ for scli, sdc/ for drv_cfg, and so forth.
  Command name (scli, drv_cfg, and so forth) is stripped. The first meaningful argument becomes the filename. Files are assigned as the relevant suffix, .txt by default.

Examples: 

• • mdm/query_cluster.txt - output of scli --query_cluster
  • mdm/tgt_dump.txt - output of scli --debug_action --tgt_dump
  • sdc/query_mdms.txt - output of drv_cfg --query_mdms
  • sds/sdbg.txt - output of SDBG dumpallscreens for SDS
• Product component files (as opposed to command outputs), <component>/cfg, <component>/logs, and so forth
  Copied from the component’s directory with the prefix stripped.

Examples: 

• • mdm/cfg/conf.txt - copy of /opt/emc/scaleio/mdm/cfg/conf.txt
  • sds/logs/trc.0 - copy of /opt/emc/scaleio/sds/logs/trc.0
• Host file-system files are placed at their file-system path relative to the bundle root.

Examples:

• • etc/os-release - copy of /etc/os-release
  • var/log/messages - copy of /var/log/messages
  • proc/cpuinfo - copy of /proc/cpuinfo
• Diagnostic collector (diag_coll) files are copied with the /opt prefix stripped, preserving internal structure.
  Example: diag_coll/logs/sar.0 - copy of /opt/diag_coll/logs/sar.0
• Hidden files (dot-prefixed) are "unhidden" by removing the leading dot.
• Utility execution log, get_info_run.log, placed directly under the <hostname>/ root

Bundle directory tree structure:

<hostname>/
|-- get_info_run.log                   Utility execution log
|-- server/                            General command output directory
|   |-- ip_-s_addr.txt
|   |-- uptime.txt
|   |-- uname_-a.txt
|   |-- ps_-elF.txt
|   |-- dmesg.txt
|   +-- ...                            (one file per collected command)
|
|-- mdm/                               PowerFlex component data (if installed)
|   |-- cfg/                           Configuration files (excl. PEM)
|   |-- logs/                          Trace and log files
|   |-- rep/                           Repository (if --mdm-repository)
|   |-- query_all.txt                  SCLI query outputs
|   |-- sdbg.txt                       SDBG screen dumps
|   +-- ...
|-- sds/
|-- pds/  
|-- dgwt/
|-- sdr/
|-- sdt/
|-- lia/
|-- sdc/
|-- gateway/
|
|-- diag_coll/                         Diagnostic data collector (if installed)
|   |-- logs/
|   |-- cfg/
|   +-- ...
|
|-- etc/                               Host files
|   |-- os-release
|   |-- sysconfig/
|   |-- network/
|   +-- ...
|-- var/
|   |-- log/
|   |   |-- messages
|   |   +-- ...
|   +-- ...
|-- proc/
|   |-- cpuinfo
|   |-- meminfo
|   +-- ...
|-- sys/
|-- ...
|
|-- scaleio-getinfo-extra/        Extra diagnostic data (if present)
+-- scaleio-getinfo-backup/       Backed-up configuration files (if any)

PRODUCT LOG AND CORE FILE FILTERING

The options described in this section control how product log files (also called trace files, for example trc.0, trc.1, exp.0) and core dump files are selected for inclusion in the collected bundle. They do so by defining a reference time, a time window around it, and count limits.

When no filtering options are specified, all product log files and (if core collection is enabled) all core dump files are collected. The filtering options progressively narrow this selection as described below.

Reference Time

A reference time can be set using either -E/--event-time or -C/--reference-core-file.

If neither --event-time nor --reference-core-file is given, no time window filtering is performed: the reference time defaults to the current time and is used only for proximity-based ordering when a count limit (-m or -k) is in effect (see Count Limits below).

If both -E and -C appear, the last item on the command line takes effect.

Time Window

When a reference time is set (using --event-time or --reference-core-file), a time window is established around it. The time window scope can be set using -t/--minutes-before-event and or -T/--minutes-after-event, which default to 15 and 5 minutes, respectively. Only files whose content overlaps with this window are eligible for collection.

For example, -E "2020-03-20 14:30" -t 10 -T 3 collects files covering the period 14:20:00 through 14:33:00.

--minutes-before-event and --minutes-after-event are ignored when neither --event-time nor --reference-core-file is specified.

Count Limits

A file count limit can be set using -m/--max-traces and -k/--max-cores, for log files and core files, respectively. The limit is measured per component.

When more files than NUM fall within the time window (or are available, if no window is active), the NUM files closest to the reference time are collected.

When a count limit is used without --event-time or --reference-core-file, all files are candidates (no time window) and the NUM most recent files are selected.

Filtering Logic

File filtering applies to the time window first, then the count limit:

1. Establish candidates. All product log files and or core dump files for a component are enumerated.
2. Derive content period. Product log files content represents a period. The content period is considered to start at its predecessor's last modification time (mtime), or the UNIX epoch, when no predecessor exists; it ends at the file's own mtime. Core dump files represent a point in time, at the file's mtime.
3. Apply time window (if -E or -C specified). Files whose content falls entirely outside the window are discarded from selection. For product log files, if no file falls within the window, the single file closest to the window is retained so that the bundle is never empty for a component. For core dump files, no such fallback applies.
4. Apply count limit (if -m and or -k specified). Among the remaining files, at most NUM are selected, preferring those closest to the reference time. Files before and after the reference time compete equally for selection.

AUTHENTICATION

The utility attempts to log in to the local MDM if a primary MDM process is detected listening on the expected port (default: 6611)

MDM login options are passed to the SCLI --login command and are processed by it.

If login fails, the utility terminates with an error.

When login is skipped, SCLI commands are still attempted (to support scenarios where the user has logged in manually beforehand). After three SCLI failures, a warning is displayed, and all further SCLI commands are skipped.

Login is skipped when:

• No primary MDM process is found on the local host.
• The --skip-mdm-login option is specified.

Login fails when:

• The MDM process owner is not in the authorized users list (default: root) and secure login is enabled.
• The SCLI --login command returns an error (for example, wrong credentials).

DISK SPACE

Disk space requirements for temporary files and the resulting bundle can vary considerably.

The utility attempts to minimize temporary space usage; it is limited to command outputs and copies of collected virtual file system (/proc and /sys) files.

The bundle can be streamed from a remote host with --output-file=- to minimize disk space usage on the PowerFlex host. The bundle file is written directly to standard output (stdout) when streaming; it is not created on disk.

The utility estimates the required disk space for both the temporary work directory and the output bundle before collecting data.

If the estimated required space exceeds the available space on the relevant file systems, the utility terminates with an error. This check can be bypassed with --skip-space-check.

The work directory and the output directory may reside on different file systems; each is checked independently.

The estimated space requirements are written to the utility's log file, get_info_run.log.

EXIT STATUS

FILES

ENVIRONMENT

Prerequisites

• The utility must be run as root (or a user with sufficient privileges to read component files, run diagnostic commands, and access /proc, /sys, and so forth).
• Standard utilities: tar, gzip, stat, find, awk, sed, getopt(1) (enhanced), nice.
• Optional: zip (for --zip), xz (for --xz), gdb/gcore (for --generate-cores)

Concurrency

Only one instance of get_info.sh may run at a time. The utility checks for an existing running instance using pidof(1) and terminates if an instance is found.

Signal Handling

The utility traps INT, EXIT, and TERM signals during data collection. Upon receiving a signal, it:

1. Restores any backed-up configuration files (for example, core generation settings).
2. Cleans up temporary directories
3. Exits with status 1

The execution log is preserved and its path is printed to standard error.

Examples

Collect a standard support bundle:

get_info.sh

Stream a bundle over an SSH connection, without creating a bundle file on the remote PowerFlex host:

ssh <host> 'get_info.sh --output-file=-' > getInfoDump-<host>.tgz

Use a different work directory to avoid filling up /tmp:

get_info.sh --work-dir=/var/tmp

Include custom paths in the bundle:

get_info.sh --collect-path=/opt/custom/app/logs --collect-path='/var/log/app*'

Collect the latest core dump only for the SDS and MDM components:

get_info.sh --collect-cores='mdm sds' --max-cores=1

Collect data centered around a core file's modification time, with a custom time window:

get_info.sh --reference-core-file=/opt/emc/scaleio/sds/bin/core.1000 \
            --minutes-before-event=10 \  
            --minutes-after-event=2