* Add lost functionality to Juniper mempool polling
* Update test data
* Test data for a more recent device
* Change simple string match to case insentive regex
* Update test data - CPU was affected too
* mempools to modern module
quick hacky hrstorage port
* port ucd-snmp-mib to Mempools
* Populate DB for ucd
Prep for yaml
* initial yaml attempt
* more complex conversions
fixes to YamlDiscovery, make leading $ optional and allow mib::oid format
* walk full tables and skip values
normalize percentages above 100
* handle precent only ones (specify total as 100)
* Move default polling out of YamlMempoolsDiscovery
* fixes
* Update test data hrstorage should be correct.
* perc_warn for hrstorage
* Host Resources, record buffer, cached, and shared
* Host Resources is always better, don't do both HR and UCD
* fix unix, include warning levels
* variable size
* consolidate skip_values
* define mempools schema
* number instead of integer
* more schema refactor
* one more skip_values reference
* throw error for invalid oid translation
aos6
* a* and Cisco
* updated test data
* update almost all hrstorage data files
* b*
* c* with test data
use standard cache for hrStorage
* use cache for storage module too
* hand bsnmp properly
* bdcom
* exclude total oid from yaml so it is not polled
May add a way to ignore this behavior and poll it, but I don't know if that is needed.
* automatically handle percent only values
* ciscowlc
* only poll used or free if we have used, free, and total.
* fix skipping
* the dlinkoning
fix find value when value "name" is numeric
* support numeric oids
* dnos/ftos attempt
* I guess we can't filter on total > 0
* edgecos
* e*
* f WIP
* f*
* gwd (aka g*)
* h* + procurve
* i*
* j*
* m*
* support 0% used memory (however unlikely)
* n*
* CISCO-PROCESS-MIB memory, share cache with processors module
* ignore mempools with invalid total
* p*
* quanta
* r*
fix raisecom mibs terribly broken
* s-z
* style fixes
* Move VRP back to PHP and make it actually work
* fix zynos
* update schema
* Update Cisco processor data for description bug fixes
* fix comware processors
* comware mempools with memory size
default precision to 1
* sophos-xg updated data
* hrstorage use ram size for buffers, cache, and shared
* Show memory available instead of free in device overview
* UCD, use same rrd format, store available instead of free in the db.
* Calculate availability for HOST-RESOURCES-MIB
* Convert UCD to standard polling
* rename old rrd files
* initial graph work
* graph WIP
* Graph looking decent
* Graph looking decent for hr
* remove old ucd_graph code
* handle availability mempool
more graph cleanup
* color adjustments
* remove accidental free calculation
* Update test data and fix corner cases
* fis pfsense
* update schema
* add default value for mempool_class
* fix whitespace
* update schema
* update schema correctly
* one more time
* fortigate_1500d-sensors missing oids
* Update docs.
* fix indent
* add implements MempoolsDiscovery explicitly to OS
* remove ucd_memory graph references
remove unused device_memory graph
* remove unused functions
* set devices with mempools to rediscover to prevent/minimize gaps
* use a subquery
* add overview graph
* port health mempools table
* Update device mempool
* only show overview if multiple
* Don't override user set warn percentages in discovery
* fix missed usage
* fix style
* Safety check to not rename rrd files incorrectly if migration has not been run.
* Fix overview percent bar and represent available and used on the bar
* missed an item to convert to mempool_class
* percent on the wrong side
* Fix routing engine discovery on standalone JunOS devices
In #9426, the sensor discover code was replaced with YAML.
Unfortunately it treats the FRU table as standalone OIDs, and did not
implement the filtering that was present in the original code.
This change causes LibreNMS to fully walk the FRU table, which allows us
to silence alerting on devices with an 'unknown' status. Without this
change, standalone devices have a spurious routing engine with state 1,
which triggers false harware failure alerts.
As we're changing the name of the sensors, this causes a loss of sensor
history.
This has been brought up on community a couple of times:
https://community.librenms.org/t/juniper-qfx-routing-engine-state-alert/6152https://community.librenms.org/t/spurious-failed-routing-engine-from-some-junos-devices/6208
* Update test data
* Update test data
* Update test data
* Update test data
* This syntax works, but fails linting
* add ability to discover percent warning on mempools discovery
* remove wierd charachters from json
* remove database defaults, add defaults for Linux and Vmware
* update db schema
* update test data
* adapt hrstorage mempools
* rebase
* Default to null for group yaml discovery.
* Update test data for a154bda yaml group null fix.
* Changes to guessed limit functions for sensors.
Original behaviour
===================
The file `includes/discovery/functions.inc.php` contains
`sensor_limit_low()` and `sensor_limit()` which both attempt to
guess a sane value for sensors when no explicitly defined
low_limit or high_limit can be found during discovery.
Both switch control structures used in those two functions
have empty case statements which means that if one of those matches,
it's going to fall through and run the code for each subsequent
case until a `break` is reached.
For example, when we call `function sensor_low_limit(dbm, -13.036)`
it will return the value `-12.3842` instead of `null`. That is
because there will be a match at `case 'dbm':` which falls through
all the way to `case 'cooling':`, where it performs
`$limit = -13.036 * 0.95` before hitting a `break`.
Changed behaviour
===================
Removed `power_consumed` and `count` guessed low_limit and
high_limit, I personally added those sensor classes in PR #9471
when I didn't understand that a switch control structure has
fall-through behaviour so I can guarantee that guessing limits for
those is a mistake on my behalf. It should not be there, only
power_factor can have guessed limits. Apologies for the issue,
I'm still a beginning programmer!
Furthermore, I removed guessed high_limit values for `current`
and `power` because these are supposed to draw higher values as
more devices or components are installed on for example a PDU or
a chassis.
Finally, I removed guessed low_limit and high_limit for `dbm`
sensors, there is a much too large variance in power budget on
commercially available optical transceivers for there to be a
sensible window where you can guess these values.
* Documentation on adding sensor classes.
* Update test data - sensor limit changes @ 30212d2
