diff --git a/ChangeLog.md b/ChangeLog.md --- a/ChangeLog.md +++ b/ChangeLog.md @@ -1,3 +1,15 @@ +# 0.1.1.0 -- 2019-05-23 + +* Change all use of `Int` to `Int64`. `Int` is only guaranteed to be able to + represent values up to 2^29-1, but entries in /proc/stat can often exceed that + (for example processes with a high resident memory usage) on 32-bit machines. + +* Export `process_open_fds`, a count of the number of entries in + `/proc/$PID/fd`. + +* Export `process_max_fds`, which exports the soft ulimit of the maximum number + of file descriptors this process may open. + # 0.1.0.0 -- 2019-01-15 * Initial release. diff --git a/prometheus-proc.cabal b/prometheus-proc.cabal --- a/prometheus-proc.cabal +++ b/prometheus-proc.cabal @@ -1,5 +1,5 @@ name: prometheus-proc -version: 0.1.0.0 +version: 0.1.1.0 synopsis: Export metrics from /proc for the current process homepage: https://github.com/fimad/prometheus-haskell license: BSD3 @@ -18,10 +18,12 @@ library exposed-modules: Prometheus.Metric.Proc build-depends: base >=4.10 && <4.13 + , directory >= 1.2.5.0 && < 1.4 , filepath >=1.4 && <1.5 - , unix-memory >=0.1 && <0.2 - , unix >=2.7 && <2.8 - , regex-applicative >=0.3 && <0.4 , prometheus-client >= 1.0.0 && < 1.1 + , regex-applicative >=0.3 && <0.4 + , text >= 0.7 && < 1.3 + , unix >=2.7 && <2.8 + , unix-memory >=0.1 && <0.2 hs-source-dirs: src default-language: Haskell2010 diff --git a/src/Prometheus/Metric/Proc.hs b/src/Prometheus/Metric/Proc.hs --- a/src/Prometheus/Metric/Proc.hs +++ b/src/Prometheus/Metric/Proc.hs @@ -2,6 +2,7 @@ {-# language NamedFieldPuns #-} {-# language OverloadedStrings #-} {-# language RecordWildCards #-} +{-# language ViewPatterns #-} {-| @@ -12,14 +13,21 @@ module Prometheus.Metric.Proc ( ProcMetrics(..), procMetrics ) where import Data.Char ( isSpace ) -import Data.Maybe ( catMaybes ) +import Data.Int ( Int64 ) +import Data.List ( isPrefixOf ) +import Data.Maybe ( catMaybes, maybeToList ) import Data.String ( fromString ) +import Data.Text ( Text, unpack ) +import Data.Text.IO ( readFile ) import Foreign.C +import Prelude hiding ( readFile ) import Prometheus +import System.Directory ( listDirectory ) import System.FilePath import System.IO.Unsafe import System.Posix.Memory ( sysconfPageSize ) import System.Posix.Process ( getProcessID ) +import System.Posix.Types ( ProcessID ) import qualified Text.Regex.Applicative as RE import qualified Text.Regex.Applicative.Common as RE @@ -62,14 +70,56 @@ getProcessID mprocStat <- - RE.match parseProcStat <$> readFile ( "/" "proc" show pid "stat" ) + RE.match parseProcStat . unpack <$> readFile ( procPidDir pid "stat" ) + processOpenFds <- + collectProcessOpenFds pid + + processMaxFds <- + collectProcessMaxFds pid + return - ( foldMap ( toMetrics ) mprocStat ) + ( [ processOpenFds ] + <> maybeToList processMaxFds + <> foldMap ( procStatToMetrics ) mprocStat + ) -toMetrics :: ProcStat -> [ SampleGroup ] -toMetrics ProcStat{ utime, stime, starttime, vsize, rss } = +collectProcessOpenFds :: ProcessID -> IO SampleGroup +collectProcessOpenFds pid = do + fmap + ( metric "process_open_fds" "Number of open file descriptors." GaugeType . length ) + ( listDirectory ( procPidDir pid "fd" ) ) + + +collectProcessMaxFds :: ProcessID -> IO ( Maybe SampleGroup ) +collectProcessMaxFds pid = do + limitLines <- + lines . unpack <$> readFile ( procPidDir pid "limits" ) + + case filter ( "Max open files" `isPrefixOf` ) limitLines of + ( words -> _max : _open : _files : n : _ ) : _ -> + return + ( Just + ( metric + "process_max_fds" + "Maximum number of open file descriptors." + GaugeType + ( read n :: Int ) + ) + ) + + _ -> + return Nothing + + +procPidDir :: ProcessID -> FilePath +procPidDir pid = + "/" "proc" show pid + + +procStatToMetrics :: ProcStat -> [ SampleGroup ] +procStatToMetrics ProcStat{ utime, stime, starttime, vsize, rss } = catMaybes [ Just process_cpu_seconds_total , process_start_time_seconds @@ -110,21 +160,23 @@ "process_resident_memory_bytes" "Resident memory size in bytes." GaugeType - ( rss * sysconfPageSize ) + ( rss * fromIntegral sysconfPageSize ) - metric metricName metricHelp metricType value = - SampleGroup - Info{..} - metricType - [ Sample - metricName - [] - ( fromString ( show value ) ) - ] +metric :: Show a => Text -> Text -> SampleType -> a -> SampleGroup +metric metricName metricHelp metricType value = + SampleGroup + Info{..} + metricType + [ Sample + metricName + [] + ( fromString ( show value ) ) + ] + -- | Convert a number of clock ticks into the corresponding duration in seconds. -fromTicks :: Int -> Double +fromTicks :: Int64 -> Double fromTicks ticks = fromIntegral ticks / fromIntegral clk_tck @@ -141,27 +193,27 @@ -} {-# NOINLINE mbtime #-} -mbtime :: Maybe Int +mbtime :: Maybe Int64 mbtime = unsafePerformIO $ do - fmap ( \( _, a, _ ) -> a ) . RE.findFirstInfix ( "btime " *> RE.decimal ) + fmap ( \( _, a, _ ) -> a ) . RE.findFirstInfix ( "btime " *> RE.decimal ) . unpack <$> readFile "/proc/stat" -- | Specific metrics from @/proc/xyz/stat@ that we are interested in. data ProcStat = ProcStat - { utime :: Int + { utime :: Int64 -- ^ Amount of time that this process has been scheduled in user mode, -- measured in clock ticks (divide by sysconf(_SC_CLK_TCK)). - , stime :: Int + , stime :: Int64 -- ^ Amount of time that this process has been scheduled in kernel mode, -- measured in clock ticks (divide by sysconf(_SC_CLK_TCK)). - , starttime :: Int + , starttime :: Int64 -- ^ The time the process started after system boot. In kernels before Linux -- 2.6, this value was expressed in jiffies. Since Linux 2.6, the value is -- expressed in clock ticks (divide by sysconf(_SC_CLK_TCK)). - , vsize :: Int + , vsize :: Int64 -- ^ Virtual memory size in bytes. - , rss :: Int + , rss :: Int64 -- ^ Resident Set Size: number of pages the process has in real memory. This -- is just the pages which count toward text, data, or stack space. This -- does not include pages which have not been demand-loaded in, or which are