EMC PowerPath/VE 5.8 Is GA!–you want to read this one
Hi, after a long beta program we finally released PP/VE 5.8, there are some exciting features in this release, so let’s just get to it: let’s dive in.. Virtual License […]
Dell Storage, PowerStore, PowerFlex PowerMax & PowerScale, Virtualization & Containers Technologies
Hi, after a long beta program we finally released PP/VE 5.8, there are some exciting features in this release, so let’s just get to it: let’s dive in.. Virtual License […]
Hi,
after a long beta program we finally released PP/VE 5.8, there are some exciting features in this release, so let’s just get to it:
let’s dive in..
Virtual License Appliance
•VMware vApp
•Preinstalled RTOOLS and PP/VE ELMS packages
–Administration of RTOOLS and electronic licenses doesn’t change in the 5.8 version
•The distribution format for vApp is OVF with corresponding vmdk.
Our appliance is the output of VMware Studio, so it fully qualifies as a vApp. Our roadmap builds on this appliance with additional features/functionality in the future.
Path Management Insight
From its position in the I/O stack, PowerPath can:
•Provide insight
–Application I/O patterns
•Read/write ratios
•I/O sizes
–Diagnostics for I/O problems
•Detect storage subsystem problems
•Detect flaky paths
•Detect latency crossing a threshold
•Avoid performance issues
–Autostandby for flaky paths
–Autostandby for cross-connected VPLEX Metro configuration
Characterization of I/O Access Patterns
•LUN based metrics:
–Throughput: bytes/second
–Average Response Time: milliseconds
–IOs/second (verbose only)
•Four size buckets with division points at 4KB, 8KB, 128KB
LUN metrics by size bucket is unique to PowerPath
Path Diagnostic Metrics
•Latency High and Low Water Marks in milliseconds
•Latency distribution (verbose mode)
–Four buckets with division points at 1, 10 and 100 milliseconds
•Retry Delta
–Number of times I/O had to be retried on another path since last sample
• Error Delta
–Number of alive-to-dead transitions since last sample
Latency distribution is unique to PowerPath
Display Performance Monitoring
# rpowermt display perf bus [options]
•Displays a bus view of the performance information
–Summary of all paths that traverse the bus
–For every bus, latency, Retry and Error Deltas are displayed
•Does not accept “dev=<device id>” or “class= <class>”
–“Invalid option: dev=all is required for this version” error is displayed if dev=anything is supplied
–“Invalid option: class=<class> is not supported for this version” error is displayed if class=<class> is supplied
Display Performance Monitoring Options
# rpowermt display perf bus
[continuous][verbose][wide|width=<#col>]
•Continuous
–Outputs the data at the completion of each interval
•If the measured data is not yet available, display timestamp and time to wait, (e.g. “Performance data ready in <xx> seconds.”)
–Not supplied: display single most recently measured values
•If the measured data is not yet available, display time to wait (e.g. “Performance data ready in <xx> seconds.”)
•Verbose – adds to the display output…
–IO counts in the latency buckets for each IT nexus:
Y <= 1ms, 1ms < Y <= 10ms, 10ms < Y <= 100ms, Y > 100ms
•Wide – 132 character width or # of columns
AutoStandby
•Currently in PowerPath, if you do not want a path used, you can set it to standby
–A standby path could be used if it is the only available path (just weighted higher so that it is not likely to be used for normal load balancing)
•Autostandby adds a trigger that causes a path to be put in autostandby mode
–Behaves just like standby
–User set standby always has precedence
•Use cases
–Detect a flaky path
–Detect groups of paths connected to local and remote array controllers in a federation (VPLEX Metro)
Flaky Path Trigger
Uses a running average of I/Os per failure (IOPF) per path
•An I/O failure on a path is either:
–A read/write failure on a path that succeeds when retried on another path, or
–An inquiry failure on a path
•Running average is calculated over last 4 recorded errors (denominator is 3)
•Trigger fires if IOPF is greater than 1 and less than or equal to user settable limit
–Default is 6000
–Needs to be greater than 1, or it’s just a dead path
–Will trigger no matter what the rate of I/Os is
Flaky Path Trigger Example
Path tests fail intermittently:
* The flaky path trigger fires, and the path ends up in “asb:iopf” mode so that it will not be used for io after future restores until the aging period expires.
Flaky Path Trigger
•Aging period default is 7 days
•Path stays in autostandby mode throughout aging period
•After aging period expires, the path goes back to active mode
•Aging period can range from 0 (aging is disabled) to 24855 days (68 years)
Flaky Path Display Example
atsb:iopf = indicates the path was set to autostandby due to a flaky path
atsb:prox = indicates the path was set to proximity (vplex) due to proximity trigger
Path Group Proximity Trigger
•Used for VPLEX cross connected metro clusters
•Based on asymmetric latency of inquiry
–Usually calculated at device discovery, or
–When proximity autostandby is enabled, or
–When recalibrating via CLI command
•No aging
•Is restored when IOPF autostandby is cleared
Paths Display Example
atsb:iopf = indicates the path was set to autostandby due to a flaky path
atsb:prox = indicates the path was set to proximity (vplex) due to proximity trigger
Reactive Autorestore
•Periodic autorestore (currently available in PowerPath) and reactive autorestore are the modes in PowerPath by which a dead path can be brought alive
•When reactive_autorestore is on, PowerPath/VE reactively tests dead paths and, if they pass the test, restores them to service
•When reactive_autorestore is off, PowerPath/VE suppresses path testing and restoration in the context of I/O, thereby quickening I/O failure when all paths to a volume are dead.
•Use Case:
–Customer does not want a dead path to brought back to service. They wanted PowerPath to fail the IO due to unavailability of paths so that higher level software (such as clusters and host volume mirroring) can do the failover instead. PowerPath by its inherent behavior will always try its best to succeed the IO by doing retries after restoring the paths.
CLI Commands
#rpowermtset reactive_autorestore={on|off} [class=<class>|all] [force]
•rpowermtset reactive_autorestore =on
–By default on installing PowerPath on a host, reactive auto restore is always ON
–PowerPath will restore the paths and try bring the paths to an active state if there are no available paths to service the IO when it is dispatched
•rpowermtset reactive_autorestore = off
–PowerPath multipathing software will avoid doing an auto restore
–PowerPath will fail the IO immediately if there are no paths available in active state to service the IO
the release notes are:
Product Overview
PowerPath is a server-resident software solution designed to enhance performance and application availability. PowerPath combines automatic load balancing, path failover, multiple path I/O capabilities, and recover functionality into one integrated package. PowerPath supports servers, including cluster servers, connected to EMC and third-party arrays.
Information about related PowerPath technologies appears in the PowerPath/VE for VMware vSphere Release Notes.
New Feature Summary
PowerPath/VE 5.8 for VMware includes the following new multipathing features: