fche changed the topic of #systemtap to: http://sourceware.org/systemtap; email systemtap@sourceware.org if answers here not timely, conversations may be logged
hpt has joined #systemtap
modem_ has quit [Read error: Connection reset by peer]
pfalleno1 has quit [Ping timeout: 245 seconds]
khaled has joined #systemtap
scox has joined #systemtap
pfallenop has joined #systemtap
khaled has quit [Quit: Leaving]
pfallenop has quit [Ping timeout: 250 seconds]
pfallenop has joined #systemtap
pfallenop has quit [Ping timeout: 250 seconds]
pfallenop has joined #systemtap
pfallenop has joined #systemtap
pfallenop has quit [Remote host closed the connection]
pfallenop has joined #systemtap
irker895 has quit [Quit: transmission timeout]
ego_ has joined #systemtap
ego_ has quit [Ping timeout: 240 seconds]
ego_ has joined #systemtap
ego_ has quit [Ping timeout: 276 seconds]
ego_ has joined #systemtap
hkshaw has joined #systemtap
ego_ has quit [Ping timeout: 260 seconds]
ego_ has joined #systemtap
srikar_away is now known as srikar
naveen has joined #systemtap
ego_ has quit [Ping timeout: 260 seconds]
hkshaw has quit [Quit: Leaving.]
pfallenop has quit [Ping timeout: 248 seconds]
hkshaw has joined #systemtap
hkshaw has quit [Client Quit]
hkshaw has joined #systemtap
ananth has joined #systemtap
pfallenop has joined #systemtap
pfallenop has quit [Ping timeout: 256 seconds]
pfallenop has joined #systemtap
pfallenop has quit [Remote host closed the connection]
pfallenop has joined #systemtap
prerna has joined #systemtap
prerna has quit [Quit: This computer has gone to sleep]
ego has joined #systemtap
ego has quit [Ping timeout: 264 seconds]
ego has joined #systemtap
prerna has joined #systemtap
hpt has quit [Ping timeout: 272 seconds]
mjw has quit [Quit: Leaving]
mjw has joined #systemtap
ego has quit [Ping timeout: 256 seconds]
nkambo has quit [Ping timeout: 245 seconds]
modem_ has joined #systemtap
ego has joined #systemtap
nkambo has joined #systemtap
naveen has quit [Quit: Leaving.]
nkambo has quit [Ping timeout: 256 seconds]
scox has quit [Ping timeout: 240 seconds]
zodbot has quit [Disconnected by services]
hkshaw has quit [Ping timeout: 264 seconds]
modem has quit [Ping timeout: 264 seconds]
pmuldoon has quit [Ping timeout: 264 seconds]
zodbot has joined #systemtap
pmuldoon has joined #systemtap
modem has joined #systemtap
hkshaw has joined #systemtap
hkshaw has quit [Client Quit]
ananth has quit [Quit: Leaving]
nkambo has joined #systemtap
nkambo1 has joined #systemtap
nkambo has quit [Ping timeout: 248 seconds]
nkambo2 has joined #systemtap
nkambo1 has quit [Ping timeout: 252 seconds]
nkambo2 has quit [Ping timeout: 256 seconds]
hkshaw has joined #systemtap
nkambo has joined #systemtap
nkambo has quit [Ping timeout: 245 seconds]
nkambo has joined #systemtap
hkshaw has quit [Ping timeout: 240 seconds]
flu_ has joined #systemtap
nkambo has quit [Ping timeout: 252 seconds]
nkambo has joined #systemtap
nkambo has quit [Ping timeout: 240 seconds]
nkambo has joined #systemtap
nkambo1 has joined #systemtap
nkambo has quit [Ping timeout: 256 seconds]
hkshaw has joined #systemtap
nkambo2 has joined #systemtap
nkambo1 has quit [Ping timeout: 276 seconds]
prerna has quit [Quit: This computer has gone to sleep]
wcohen has quit [Ping timeout: 245 seconds]
hkshaw has quit [Quit: Leaving.]
scox has joined #systemtap
nkambo2 has quit [Ping timeout: 240 seconds]
nkambo2 has joined #systemtap
mbenitez has joined #systemtap
mbenitez has quit [Changing host]
mbenitez has joined #systemtap
ego has quit [Ping timeout: 248 seconds]
brolley has joined #systemtap
tromey has joined #systemtap
srikar is now known as srikar_away
srikar_away is now known as srikar
wcohen has joined #systemtap
nkambo1 has joined #systemtap
nkambo2 has quit [Ping timeout: 240 seconds]
sfink has quit [Ping timeout: 245 seconds]
nkambo1 has quit [Ping timeout: 260 seconds]
nkambo has joined #systemtap
jameshyde_ has joined #systemtap
<jameshyde_> is it possible to use systemtap to see activities for pgfree/s in sar?
eichiro has joined #systemtap
scox has quit [Ping timeout: 264 seconds]
<fche> jameshyde_, most likely
<fche> just need to reverse engineer what /proc value sar is pulling out for that field, and what kernel-side event that corresponds to
<fche> there is kernel.trace("kmem:mm_page_free") and kernel.trace("kmem:mm_page_free_batched") e.g.
<jameshyde_> ok thanks
<jameshyde_> fche: when trying this in RHEL6, got this:
<jameshyde_> source: probe kernel.trace("kmem:mm_page_free") {
<jameshyde_> semantic error: no match (similar tracepoints: kfree, mm_pagevec_free, kmem_cache_free, mm_page_alloc, module_free)
<fche> ah yes, tracepoints do vary between kernels
<fche> am just hunting down some info for you now, stand by
<fche> (sar pgfree/s comes from /proc/vmstat line pgfree)
<jameshyde_> ok
<jameshyde_> we noticed large number of pgfree/s, so wondering which proc is really doing that. systemtap might be able to pinpoint this..
<fche> should be able to, definitely
zorachus has left #systemtap [#systemtap]
<fche> ogling linux mm/vmstat.c , include/linux/vmstat.h, include/linux/vm_event_item.h
<fche> aaaand the winner is __count_vm_event(PGFREE);
<fche> inconveniently, I don't see kernel tracepoints in the vicinity, in the upstream git kernel
<fche> jameshyde_, just wondering for testing purposes, what version of rhel6 are you running?
<fche> jameshyde_, ok, have something for you to try:
<jameshyde_> rhel6.6 here
<fche> these are approximations only - may overcount the pgfree/s metric a little bit
<fche> they work by tracing the entry to both functions in mm/* that can cause incrementing of the PGFREE counter
<fche> now if you want to track -rates-, for example rates in excess of N pages per second, we can do that
<jameshyde_> thanks fche, this is very helpful! would love to have that if possible ;)
<fche> ok, will do a v2 moohahha
<fche> this one periodically reports accumulated pgfree counts on a per-thread basis
naveen has joined #systemtap
<jameshyde_> awesome! this is great, thanks a lot.
prerna has joined #systemtap
<fche> ok. can you work from that, or shall we try generalizing it further?
<fche> we already have the eventcount.stp script that could be used like ...
<fche> stap eventcount.stp 'kernel.function("free_hot_cold_page")' 'kernel.function("__free_pages_ok")'
<fche> except that can't account for $order (power-of-two sets of pages)
prerna has quit [Quit: This computer has gone to sleep]
<jameshyde_> fche: would love to have it generalized a bit further ;)
<fche> what would you like ideally?
srikar is now known as srikar_away
naveen has quit [Quit: Leaving.]
<fche> jameshyde_, ^^ ?
<jameshyde_> fche: is it possible to seperate kernel proc from user proc?
<jameshyde_> ppid=2 would be kernel proc
<jameshyde_> this is pretty good already ;)
<fche> thing is some of these paging decisions may be performed -by- kernel threads, but on behalf of userspace actions
sfink has joined #systemtap
<fche> (in a sense they are all userspace actions)
<fche> it's almost as we'd like to know whether the target page was mapped into a userspace process or not
<jameshyde_> ok
<fche> There are $page->flags one could look at, but it's getting pretty far from my expertise
<fche> if you know that you're getting high numbers (rates) from some particular process,
<fche> a version of the second script could do something like
<fche> function freed(n) { history ... <<< n; if (@sum(history[...]) > NUMBER) { print_backtrace() } ' kind of thing
<fche> ie look back up the callstack to see why
<fche> print_ubacktrace() too possibly
<jameshyde_> this might be exactly what we need, thanks!
<fche> it's an iterative process to discover what's going on; can't give you an all-singing all-question-answering script
dancancode has joined #systemtap
scox has joined #systemtap
modem_ has quit [Ping timeout: 240 seconds]
modem_ has joined #systemtap
flu_ has quit [Quit: flu_]
nkambo1 has joined #systemtap
nkambo has quit [Ping timeout: 260 seconds]
nkambo2 has joined #systemtap
nkambo1 has quit [Ping timeout: 252 seconds]
nkambo1 has joined #systemtap
nkambo2 has quit [Ping timeout: 240 seconds]
nkambo1 has quit [Ping timeout: 272 seconds]
nkambo has joined #systemtap
nkambo1 has joined #systemtap
nkambo has quit [Ping timeout: 245 seconds]
nkambo1 has quit [Ping timeout: 240 seconds]
nkambo1 has joined #systemtap
nkambo2 has joined #systemtap
nkambo1 has quit [Ping timeout: 240 seconds]
mbenitez has quit [Quit: Leaving]
tromey has quit [Quit: ERC (IRC client for Emacs 25.0.90.1)]
nkambo1 has joined #systemtap
nkambo2 has quit [Ping timeout: 272 seconds]
nkambo2 has joined #systemtap
nkambo1 has quit [Ping timeout: 245 seconds]
nkambo1 has joined #systemtap
nkambo2 has quit [Ping timeout: 260 seconds]
nkambo2 has joined #systemtap
nkambo1 has quit [Ping timeout: 276 seconds]
nkambo has joined #systemtap
nkambo2 has quit [Ping timeout: 245 seconds]
nkambo has quit [Read error: Connection reset by peer]
wcohen has quit [Ping timeout: 240 seconds]
nkambo has joined #systemtap
sfink has quit [Ping timeout: 240 seconds]
brolley has left #systemtap [#systemtap]
nkambo has quit [Ping timeout: 240 seconds]
nkambo1 has joined #systemtap
nkambo1 has quit [Ping timeout: 272 seconds]
wcohen has joined #systemtap
mjw has quit [Quit: Leaving]