haproxy <- read.csv("some.csv")
options(digits.secs = 3)
haproxy$date<-as.POSIXct(haproxy$timestamp-28800,origin="1970-01-01",tz="UTC")
head(haproxy,10)
# find out qps , count of records per time series
# per second split
haproxy_qps <- data.frame(table(cut(haproxy_date, breaks="sec")))
# plot it
plot(haproxy_qps,type="l")
# filter data based on filed value
haproxy_pl <- subset(haproxy,type=='idx')
# plot qps
plot(table(cut(haproxy_ts$date, breaks="sec")),type="l")
# uniq values in col
unique(haproxy$pid)
# loop through pids
for (p in unique(haproxy$pid) ) { print (summary(subset(haproxy,pid==p)$type) ) }
#doing math of col vals and adding new col with results
haproxy$req_lat <- (haproxy$Tt - haproxy$Ti)
plot(haproxy$date,haproxy$req_lat,type="l")
# percentage of requests with various HTTP responses
x <- data.frame(table(hp_before$status))
# add column names
colnames(x) <- c('status','freq')
# do the math
x$perc <- x$freq / sum(x$freq) *100
# rename existing column name
library(data.table)
setnames(d, old = c('a','d'), new = c('anew','dnew'))
# various quantiles
with_numa_Ta = with_numa$Ta
quantile(with_numa_Ta, c(.10,.50,.75,.90,.99,.999))
# 10% 50% 75% 90% 99% 99.9%
# 4 7 14 265 398 1577