memory alloc and probe

MAGNIEN, Thierry thierry.magnien at sfr.com
Thu Jun 19 08:07:56 CEST 2014


Hi,

1) Tou have to check if you have objects that are removed from cache because of lack of space. They are marked as "n_lru_nuked" in varnishstat.

2) With your probe configuration, varnish will have to get at least 8 good results out of 10 from probes in order to consider the backend healthy. With an interval of 30s, this can take quite long ;-)

Regards,
Thierry



-----------------------------
De : varnish-misc-bounces+thierry.magnien=sfr.com at varnish-cache.org [mailto:varnish-misc-bounces+thierry.magnien=sfr.com at varnish-cache.org] De la part de Hernán Marsili
Envoyé : mercredi 18 juin 2014 19:15
À : varnish-misc
Objet : memory alloc and probe

Hi,

We have two problems while using Varnish:

1) we actually don't know how to determine if the amount of memory we assigned to the MALLOC is enough or not. varnishstat -1 show this:
 
SMA.s0.c_req               58035        34.08 Allocator requests
SMA.s0.c_fail                  0         0.00 Allocator failures
SMA.s0.c_bytes        4830340101   2836371.17 Bytes allocated
SMA.s0.c_freed        2864339421   1681937.42 Bytes freed
SMA.s0.g_alloc             44952          .   Allocations outstanding
SMA.s0.g_bytes        1966000680          .   Bytes outstanding
SMA.s0.g_space         181482968          .   Bytes available
SMA.Transient.c_req        72960        42.84 Allocator requests
SMA.Transient.c_fail           0         0.00 Allocator failures
SMA.Transient.c_bytes   4366160011   2563805.06 Bytes allocated
SMA.Transient.c_freed   4365783718   2563584.10 Bytes freed
SMA.Transient.g_alloc          350          .   Allocations outstanding
SMA.Transient.g_bytes       376293          .   Bytes outstanding
SMA.Transient.g_space            0          .   Bytes available
2) PROBE: we configured the probes as follows. What we noticed is, once the backend is mark as sick, even when the service is restored, Varnish continues to answering with a GURU MEDITATION. Is like VARNISH never checks back again the backend to see if it is healthy again. Is that a misconfiguration we have?

 .probe = {
      .url = "http://www.tycsports.com/balancer/alive1.html";
      .interval = 30s;   #chequear cada X seg
      .timeout = 10s;    #timeout de la prueba
      .window = 10;      #el probing funciona como sliding window
      .threshold = 8;   # cuantos del window tienen que estar ok para declararl$
 
Any help will be much appreciated! tnx :)

Regards,
Hernán.


More information about the varnish-misc mailing list