Linux fields
linux module
linux system metrics
conntrack
summary of nf_conntrack statistics, summed across CPU cores
linux.conntrack.summary.drop- packets dropped due to conntrack failiure
type: long
linux.conntrack.summary.early_drop- conntrack entries dropped to make room for new ones
type: long
linux.conntrack.summary.entries- entries in the conntrack table
type: long
linux.conntrack.summary.found- successfully searched entries
type: long
linux.conntrack.summary.ignore- packets seen already connected to a conntrack entry
type: long
linux.conntrack.summary.insert_failed- Number of entries where list insert insert failed
type: long
linux.conntrack.summary.invalid- packets seen that cannot be tracked
type: long
linux.conntrack.summary.search_restart- table lookups which had to be restarted due to table resizes
type: long
iostat
linux.iostat.read.request.merges_per_sec- The number of read requests merged per second that were queued to the device.
type: float
linux.iostat.write.request.merges_per_sec- The number of write requests merged per second that were queued to the device.
type: float
linux.iostat.read.request.per_sec- The number of read requests that were issued to the device per second
type: float
linux.iostat.write.request.per_sec- The number of write requests that were issued to the device per second
type: float
linux.iostat.read.per_sec.bytes- The number of Bytes read from the device per second.
type: float
format: bytes
linux.iostat.read.await- The average time spent for read requests issued to the device to be served.
type: float
linux.iostat.write.per_sec.bytes- The number of Bytes write from the device per second.
type: float
format: bytes
linux.iostat.write.await- The average time spent for write requests issued to the device to be served.
type: float
linux.iostat.request.avg_size- The average size (in bytes) of the requests that were issued to the device.
type: float
linux.iostat.queue.avg_size- The average queue length of the requests that were issued to the device.
type: float
linux.iostat.await- The average time spent for requests issued to the device to be served.
type: float
linux.iostat.service_time- The average service time (in milliseconds) for I/O requests that were issued to the device.
type: float
linux.iostat.busy- Percentage of CPU time during which I/O requests were issued to the device (bandwidth utilization for the device). Device saturation occurs when this value is close to 100%.
type: float
ksm
KSM statistics
linux.ksm.stats.pages_shared- Shared pages in use.
type: long
linux.ksm.stats.pages_sharing- Sites sharing pages.
type: long
linux.ksm.stats.pages_unshared- Unique pages.
type: long
linux.ksm.stats.full_scans- Number of times mergable pages have been scanned.
type: long
linux.ksm.stats.stable_node_chains- Pages that have reached max_page_sharing.
type: long
linux.ksm.stats.stable_node_dups- Number of duplicated KSM pages.
type: long
Linux memory data
memory page statistics
linux.memory.page_stats.pgscan_kswapd.pages- pages scanned by kswapd
type: long
format: number
linux.memory.page_stats.pgscan_direct.pages- pages scanned directly
type: long
format: number
linux.memory.page_stats.pgfree.pages- pages freed by the system
type: long
format: number
linux.memory.page_stats.pgsteal_kswapd.pages- number of pages reclaimed by kswapd
type: long
format: number
linux.memory.page_stats.pgsteal_direct.pages- number of pages reclaimed directly
type: long
format: number
linux.memory.page_stats.direct_efficiency.pct- direct reclaim efficiency percentage. A lower percentage indicates the system is struggling to reclaim memory.
type: scaled_float
format: percent
linux.memory.page_stats.kswapd_efficiency.pct- kswapd reclaim efficiency percentage. A lower percentage indicates the system is struggling to reclaim memory.
type: scaled_float
format: percent
This group contains statistics related to huge pages usage on the system.
linux.memory.hugepages.total- Number of huge pages in the pool.
type: long
format: number
linux.memory.hugepages.used.bytes- Memory used in allocated huge pages.
type: long
format: bytes
linux.memory.hugepages.used.pct- Percentage of huge pages used.
type: long
format: percent
linux.memory.hugepages.free- Number of available huge pages in the pool.
type: long
format: number
linux.memory.hugepages.reserved- Number of reserved but not allocated huge pages in the pool.
type: long
format: number
linux.memory.hugepages.surplus- Number of overcommited huge pages.
type: long
format: number
linux.memory.hugepages.default_size- Default size for huge pages.
type: long
format: bytes
This group contains statistics related to the swap memory usage on the system.
linux.memory.swap.total- Total swap memory.
type: long
format: bytes
linux.memory.swap.used.bytes- Used swap memory.
type: long
format: bytes
linux.memory.swap.free- Available swap memory.
type: long
format: bytes
linux.memory.swap.used.pct- The percentage of used swap memory.
type: scaled_float
format: percent
linux.memory.swap.in.pages- Pages swapped in.
type: long
linux.memory.swap.out.pages- Pages swapped out.
type: long
linux.memory.swap.readahead.cached- Swap readahead pages hit from swap_ra_hit.
type: long
linux.memory.swap.readahead.pages- Pages swapped based on readahead predictions.
type: long
pageinfo
Data from /proc/buddyinfo grouping used pages by order
DMA page Data
linux.pageinfo.buddy_info.DMA.0- free chunks of 2^0*PAGE_SIZE
type: long
linux.pageinfo.buddy_info.DMA.1- free chunks of 2^1*PAGE_SIZE
type: long
linux.pageinfo.buddy_info.DMA.2- free chunks of 2^2*PAGE_SIZE
type: long
linux.pageinfo.buddy_info.DMA.3- free chunks of 2^3*PAGE_SIZE
type: long
linux.pageinfo.buddy_info.DMA.4- free chunks of 2^4*PAGE_SIZE
type: long
linux.pageinfo.buddy_info.DMA.5- free chunks of 2^5*PAGE_SIZE
type: long
linux.pageinfo.buddy_info.DMA.6- free chunks of 2^6*PAGE_SIZE
type: long
linux.pageinfo.buddy_info.DMA.7- free chunks of 2^7*PAGE_SIZE
type: long
linux.pageinfo.buddy_info.DMA.8- free chunks of 2^8*PAGE_SIZE
type: long
linux.pageinfo.buddy_info.DMA.9- free chunks of 2^9*PAGE_SIZE
type: long
linux.pageinfo.buddy_info.DMA.10- free chunks of 2^10*PAGE_SIZE
type: long
linux.pageinfo.nodes.*- Raw allocation info from /proc/pagetypeinfo
type: object
Linux pressure stall information metrics for cpu, memory, and io
linux.pressure.cpu.some.10.pct- The average share of time in which at least some tasks were stalled on CPU over a ten second window.
type: float
format: percent
linux.pressure.cpu.some.60.pct- The average share of time in which at least some tasks were stalled on CPU over a sixty second window.
type: float
format: percent
linux.pressure.cpu.some.300.pct- The average share of time in which at least some tasks were stalled on CPU over a three hundred second window.
type: float
format: percent
linux.pressure.cpu.some.total.time.us- The total absolute stall time (in microseconds) in which at least some tasks were stalled on CPU.
type: long
linux.pressure.memory.some.10.pct- The average share of time in which at least some tasks were stalled on Memory over a ten second window.
type: float
format: percent
linux.pressure.memory.some.60.pct- The average share of time in which at least some tasks were stalled on Memory over a sixty second window.
type: float
format: percent
linux.pressure.memory.some.300.pct- The average share of time in which at least some tasks were stalled on Memory over a three hundred second window.
type: float
format: percent
linux.pressure.memory.some.total.time.us- The total absolute stall time (in microseconds) in which at least some tasks were stalled on memory.
type: long
linux.pressure.memory.full.10.pct- The average share of time in which in which all non-idle tasks were stalled on memory simultaneously over a ten second window.
type: float
format: percent
linux.pressure.memory.full.60.pct- The average share of time in which in which all non-idle tasks were stalled on memory simultaneously over a sixty second window.
type: float
format: percent
linux.pressure.memory.full.300.pct- The average share of time in which in which all non-idle tasks were stalled on memory simultaneously over a three hundred second window.
type: float
format: percent
linux.pressure.memory.full.total.time.us- The total absolute stall time (in microseconds) in which in which all non-idle tasks were stalled on memory.
type: long
linux.pressure.io.some.10.pct- The average share of time in which at least some tasks were stalled on io over a ten second window.
type: float
format: percent
linux.pressure.io.some.60.pct- The average share of time in which at least some tasks were stalled on io over a sixty second window.
type: float
format: percent
linux.pressure.io.some.300.pct- The average share of time in which at least some tasks were stalled on io over a three hundred second window.
type: float
format: percent
linux.pressure.io.some.total.time.us- The total absolute stall time (in microseconds) in which at least some tasks were stalled on io.
type: long
linux.pressure.io.full.10.pct- The average share of time in which in which all non-idle tasks were stalled on io simultaneously over a ten second window.
type: float
format: percent
linux.pressure.io.full.60.pct- The average share of time in which in which all non-idle tasks were stalled on io simultaneously over a sixty second window.
type: float
format: percent
linux.pressure.io.full.300.pct- The average share of time in which in which all non-idle tasks were stalled on io simultaneously over a three hundred second window.
type: float
format: percent
linux.pressure.io.full.total.time.us- The total absolute stall time (in microseconds) in which in which all non-idle tasks were stalled on io.
type: long
Wattage as reported by Intel RAPL
linux.rapl.core- The core where the RAPL request originated from. Only one core is queried per hardware CPU.
type: long
linux.rapl.dram.watts- Power usage in watts on the DRAM RAPL domain
type: float
linux.rapl.dram.joules- Raw power usage counter for the DRAM domain
type: float
linux.rapl.pp0.watts- Power usage in watts on the PP0 RAPL domain
type: float
linux.rapl.pp0.joules- Raw power usage counter for the PP0 domain
type: float
linux.rapl.pp1.watts- Power usage in watts on the PP1 RAPL domain
type: float
linux.rapl.pp1.joules- Raw power usage counter for the PP1 domain
type: float
linux.rapl.package.watts- Power usage in watts on the Package RAPL domain
type: float
linux.rapl.package.joules- Raw power usage counter for the package domain
type: float