Statistics
This page describes the statistics available in the Weka system and how to work with them.
Last updated
This page describes the statistics available in the Weka system and how to work with them.
Last updated
As the Weka system runs, it collects hundreds of statistics on system performance. These statistics help in analyzing the Weka system performance and determining the source of any problems.
Five different categories of statistics are available for review - Operations (NFS), Operations (Driver), Object Store, SSD and CPU – and when each category is selected, a list of the possible statistics that can be selected is displayed.
By default, the main statistics page displays the last hour of operation, presenting the Weka system operation per second on a time axis.
This Statistics view screen offers a number of options to drill-down into the statistics, according to category. Options include:
Mousing over the scrollable graph area to view various performance metrics of the Weka cluster.
Troubleshooting or obtaining a correlation between events and performance (using the top line which provides links to events that occurred).
Adding more statistics to the view (using the Statistics menu).
Displaying different statistics simultaneously and toggling between them. By default, the graph area shows Ops/sec for the last hour. Using the "Hour, Day, Week" buttons
at the bottom-right enables changing of the time interval.
Displaying, hiding. deleting, and zooming-in on statistics from defined timelines and dates.
Bookmarking specific statistics for future reference and sharing with others (using the URL).
Note: The statistics are only shown for backend/clients that are part of the cluster. Once a host is removed (or a client is not connected to the cluster for more than the remove_after_secs
period) its statistics will not be shown. The Weka cluster does not hold historical statistics data. Use weka-mon
for that, as suggested in the External Monitoring section.
To view the statistics screen using the GUI, click the statistics button on the left bar:
To select the addition of specific statistics, click the + Add Statistics tab on the right-hand side of the Statistics view screen. The statistics menu will be displayed.
Then select the component for which statistics required from the six possible categories. As each component is selected, the list of possible Statistics Names that can be selected changes. It is also possible to searching for a specific statistic by typing the name of the statistic in the Filter field at the top of the menu.
Up to 5 different statistics can be displayed simultaneously. Selecting a metric adds its graph to the Statistics view, together with a selector containing the category and name of the statistics which are displayed according to the appropriate units.
Switching the active unit scale is performed by clicking on one of the inactive units displayed in the left bottom corner of the graph display box.
To hide or delete statistics from a graph, mouse-over the selector and click either the Hide or Delete button appearing under the selector.
To define a specific period of time (start and end) for the statistics to be displayed, click the From and To selectors appearing in the left corner of the graph display box. Then select the date of the statistics required from the calendar popup and the hours from the right scroller, or by using the up and down arrows that appear when hovering on the time selectors.
Note: It is also possible to change the time period by dragging the graph left or right.
The Auto Refresh setting offers another option for defining a specific period of time. Activation of Auto Refresh is performed by clicking on one of the buttons displayed in the right bottom corner of the graph display box, according to the desired period (hour, day or week). The selected period of time for the statistics will be automatically updated every minute, until Auto Refresh is deactivated by clicking the 'x' button located at the top right of the Auto Refresh tool tip, or by defining a period of time using the time selector.
To obtain a summary of events that occurred in a specific time period, click the events bubble displayed above the graph. The Events popup will be displayed. Expand the popup in order to obtain a detailed list of events. Click on the icon next to each event to link to the selected event in the Events view screen.
Command: weka stats list-types
Use the following command line to obtain statistics definition information:
weka stats list-types [<name-or-category>] [--show-internal]
Parameters in Command Lines
Command: weka stats realtime
Use the following command line to obtain the current performance-related statistics of the hosts, in a one-second interval:
weka stats realtime [<node-ids>] [--raw-units] [--UTC]
Parameters in Command Lines
Command: weka stats
The collected statistics are helpful to analyze system performance and determine the source of any problems as the Weka system runs, according to several categories. When each category is selected, a list of the possible statistics that can be selected is displayed.
Note: All statistics are averaged over one-second intervals. Consequently, "total" or other aggregates relate to a specific minute.
Use the following command line to manage filters and read statistics:
weka stats [--start-time <start-time>] [--end-time <end-time>] [--interval interval] [--resolution-secs resolution-secs] [--category category][--stat stat] [--node-ids node-ids] [--param param] [--accumulated] [--per-node] [--no-zeros] [--show-internal] [--raw-units] [--UTC]
Parameters in Command Lines
Command: weka stats retention set
Use the following command line to set the statistics retention period
weka stats retention set <--days days> [--dry-run]
Parameters in Command Lines
Use weka stats retention status
to view the current retention and weka stats retention restore-default
to restore the default retention settings.
Name
Type
Value
Limitations
Mandatory
Default
name-or-category
String
Name or category fo filter by
Valid name or category
No
show-internal
Boolean
Also displays internal statistics
No
False
Name
Type
Value
Limitations
Mandatory
Default
node-ids
Comma-separated strings
Only show realtime stats of these nodes
No
raw-units
Boolean
Print values in raw units (bytes, seconds, etc.)
No
Human-readable format, e.g 1KiB 234MiB 2GiB
UTC
Boolean
Print times in UTC
No
Host's local time
Name
Type
Value
Limitations
Mandatory
Default
start-time
String
Start time of the reported period
Format: 5m, -5m, -1d, -1w, 1:00, 01:00, 18:30, 18:30:07, 2018-12-31 10:00, 2018/12/31 10:00, 2018-12-31T10:00, 9:15Z, 10:00+2:00
No
-1m
end-time
String
End time of the reported period
Format: 5m, -5m, -1d, -1w, 1:00, 01:00, 18:30, 18:30:07, 2018-12-31 10:00, 2018/12/31 10:00, 2018-12-31T10:00, 9:15Z, 10:00+2:00
No
Current time
interval
String
Period of time to be reported
Valid interval in seconds (positive integer number)
Yes
resolution-secs
String
Length of each interval in the reported period
Must be multiples of 60 seconds
No
60
category
String
Specific categories for retrieval of appropriate statistics
Valid existing categories: CPU, Object Store, Operations, Operations (NFS), Operations (Driver), SSD
No
All
stat
String
Statistics names
Valid statistics names
No
All
node-ids
String
Node id
Valid node-id
No
All
param
String
For parameterized statistics, retrieve only the instantiations where the specified parameter is of the specified value. Multiple values can be supplied for the same key, e.g. '--param method:putBlocks --param method:initBlock'
Format: key:val
No
accumulated
Boolean
Display accumulated statistics, not rate statistics
No
False
per-node
Boolean
Does not aggregate statistics across nodes
No
False
no-zeros
Boolean
Filters results where the value is 0
No
False
show-internal
Boolean
Also displays internal statistics
No
False
raw-units
Boolean
Print values in raw units (bytes, seconds, etc.)
No
Human-readable format, e.g 1KiB 234MiB 2GiB
UTC
Boolean
Print times in UTC
No
Host's local time
Name
Type
Value
Limitations
Mandatory
Default
days
Number
The Number of days to keep the statistics
Should have enough free disk space per server
Yes
dry-run
Boolean
Only test the required capacity per the retention period
No