maria-discuss team mailing list archive

[Vladislav Vaintroub <vvaintroub@xxxxxxxxx>]

On 17.07.2017 15:14, Reindl Harald wrote:
> Am 17.07.2017 um 15:02 schrieb Vladislav Vaintroub:
> > You are getting HY000/2002, the range starting with 2000 (ending iirc 
> > 3000) is the error originating on client.  If I was to guess it is a 
> > client-side connection timeout
>
> are you aware that the strace output below with "EAGAIN (Resource 
> temporarily unavailable)" is from running strace on the myslqd PID?
And what does it prove? With server doing non-blocking socket reads, 
EAGAIN is expected , and handled by switching to poll(). Since long time 
(if memory serves me right 10.0/MySQL 5.6), this is how 
net_read_timeout/net_write_timeout are handled on Unixes,on the server side.
> on a machine which spits out 4000 cms pages per second while 
> responding to additional clients via browser within 0.005 seconds 
> there souldb e not any timeout at all under a 250 concurrency
>
> > > fcntl(88, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> > > accept4(88, {sa_family=AF_UNIX}, [128->2], SOCK_CLOEXEC) = 95
> > > fcntl(88, F_SETFL, O_RDWR)              = 0
> > > fcntl(95, F_SETFD, FD_CLOEXEC)          = 0
> > > futex(0x7f65e917eae0, FUTEX_WAIT_PRIVATE, 2, NULL) = -1 EAGAIN 
> > > (Resource temporarily unavailable)
>
> _______________________________________________
> Mailing list: https://launchpad.net/~maria-discuss
> Post to     : maria-discuss@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~maria-discuss
> More help   : https://help.launchpad.net/ListHelp