Reputation
Top tag
Next privilege 5,000 Rep.
Approve tag wiki edits
Badges
4 14 37
Impact
~408k people reached

Mar
5
comment High %sys utilization in Nginx cache node
If THP is causing issues that most likely means that you are doing too many memcpy/memmove operations (you have those in your perf report). That is probably because of your lua scripts. To further improve performance of the server you can try minimizing number of memory copies in your code.
Feb
10
comment Nginx uses both writev() and sendfile()
Also TLS will require using writev since (at least for now) kernel can not encrypt data for you.
Apr
24
comment FreeBSD Established Connections Maxing out at 300
It's most likely application scalability issue not connected to kernel.
Apr
24
comment “show_signal_msg: N callbacks suppressed” then pegged at 100% and machine unresponsive
No, I just suggested that "unresponsiveness" can be connected to codedumps being written to disk.
Apr
23
comment “show_signal_msg: N callbacks suppressed” then pegged at 100% and machine unresponsive
If 12 callbacks from show_signal_msg were suppressed that means that something segfaulted 13 times in a row. Do you have coredumps enabled on that box?
Apr
17
comment kvm hypervisor kernel panic
Memory failure? What does SEL (System Event Log) says?
Mar
11
comment Centos 6.5 crashing with kernel panics
I suspect CPU failure.
Feb
14
comment freebsd 7.2 kern.threads.max_threads_per_proc doesn't change after reboot
kern.threads.max_threads_per_proc is read-write tunable so you should place it in /etc/sysctl.conf
Feb
6
comment Java causing kernel panic
It's kernel panic in interrupt context - most likely java is not a culprit there - just interrupt happen to strike in the middle of java execution.
Jan
13
comment ext4 production ready on kernel 2.6.32 (debian6)?
Bugfixes are already done by LTS (long term support) kernel maintainers. Backporting new hardware support and features is crucial for Enterprise distros, otherwise for example you would not be able to run RHEL5 on modern hardware at all. Also if RedHat would not backport features from recent kernels it will loose customers. Anyway just look at release notes of RHEL 6.5: access.redhat.com/site/documentation/en-US/… and look for has been added or support for.
Jan
13
comment ext4 production ready on kernel 2.6.32 (debian6)?
By backports I meant porting linux kernel features from recent kernels back to 2.6.32 without breaking API/ABI e.g. new versions of drivers or sometimes new features.
Jan
12
comment ext4 production ready on kernel 2.6.32 (debian6)?
It's worth mentioning that RedHat 2.6.32 has a bunch of redhat-specific backports from recent kernels, so it can not be compared to Debian's.
Jan
11
comment Linux - NTP stopped working after recent update
restrict -6 ::1?
Oct
29
comment Should I set rto_min for a route inside data center?
@Janning max latency is proportional to cumulative size of all buffers along the way. Speed of light is negligible in your case.
Oct
28
comment Should I set rto_min for a route inside data center?
I guess rto_min should be set not with regard to current latency, but w.r.t. sizes of buffers along the way and bandwidth. For example in presence of traffic bursts switch/router can hold frame/packet for couple of milliseconds.
Sep
12
comment High CPU usage in random IP/Port UDP DDoS
Oops - thanks for noticing it!
Sep
11
comment High CPU usage in random IP/Port UDP DDoS
Using ipset in blacklist mode won't help against UDP flood with random source IPs.
Aug
19
comment continous kernel EIP errors
If it's always in filesystem/journal related code then there is a change that fsck will help. PS. Also it's better to cut and paste starting from line ------------[ cut here ]------------
Jul
26
comment Join multiple filesystems (on multiple computers) into one big volume
To be taken seriously in Infiniband world ceph needs to implement native verbs interface. But still even with IPoIB on >= 12xQDR Infiniband ceph should be faster than 10G Ethernet interconnect.
Jul
12
comment How do I make OOM Killer more passive
I'm just used to it because it has -e flag that I was using when swapping on external flash drive. Here, yes I guess it's rather useless.