DNS Cache - won't flush, slows down router

RouterOS General
1

Has anyone noticed this? I’m running a number of hotspots with 2.9 (various versions) where the hotspot redirects DNS to the DNS cache. After a long period of uptime ( > 60 days) logins to the router become slow, the CPU starts to spike, the CPU idle level continues to grow up, as does the memory usage. If I click flush on the DNS cache, it only flushes some of it, as the “cache used” value will show something like 1728KB if my max is 2048KB. This only happens once the cache has reached it’s maximum. A reboot will clear the cache and memory, but the CPU utilization continues to grow.

I’ve seen this on every version of 2.9 I’ve tried (including 2.9.24) and recently upgraded one from 2.9.18 to 2.9.24 but it’s still causing CPU issues. This was also a problem with 2.8, but it was much easier to bypass the internal DNS cache as you could just disable the firewall rule that redirects it. In 2.9, even after entering bypass rules in the pre-hotspot chain, the cache continues to grow.

So, is there a way to completely avoid the DNS cache when using the Hotspot? Is anyone noticing this issue? Seems there is a permanent memory leak in the DNS caching…

2

the solve this

for example
/ip dns set cache-max-ttl=30m




3

Thanks for the reply aitsecurty. Interesting, I rarely setup DNS from the terminal other than the primary and secondary, and have actually never noticed the ttl setting in there. I’ll give this a shot and see what happens. I appreciate the suggestion :slight_smile:

Mikrotik, why is the cache-max-ttl setting not in winbox? Any chance this could be added at some point?

4

good point, i think so, Mk need make manual for winbox, and put the cache-max-ttl too in winbox too.

Test and tell what append!

5

Well, here I am a few days later…I thought originally things were back to normal. The cache size has slowly decreased to somewhere around 1MB or so (using default 2048K limit). I’ve got the ttl set to 1 hour. Today, the router started to slow down again. Here’s my cache, which is full, won’t flush, and only reports a few entries (DNS numbers have been changed):

Terminal vt102 detected, using multiline input mode
[admin@mt] > ip dns
[admin@mt] ip dns> pr
            primary-dns: 1.2.3.4
          secondary-dns: 1.2.3.5
  allow-remote-requests: no
             cache-size: 2048KiB
          cache-max-ttl: 1h
             cache-used: 2048KiB
[admin@mt] ip dns> cache
[admin@mt] ip dns cache> flush
[admin@mt] ip dns cache> ..
[admin@mt] ip dns> pr
            primary-dns: 1.2.3.4
          secondary-dns: 1.2.3.5
  allow-remote-requests: no
             cache-size: 2048KiB
          cache-max-ttl: 1h
             cache-used: 2048KiB
[admin@mt] ip dns> cache 
[admin@mt] ip dns cache> pr
Flags: S - static 
 #   NAME                                          ADDRESS         TTL         
 0   www.mycricket.com                             63.214.184.152  59m56s      
 1   ns4.customer.level3.net                       209.244.4.53    28m52s      
 2   www.jumpmobile.com                            64.78.18.76     59m57s      
[admin@mt] ip dns cache> all pr
Flags: S - static, N - negative 
 #   NAME                        TYPE  DATA                        TTL         
 0   shaharazran.com             A     69.12.119.152               59m59s      
 1   shaharazran.com             NS    ns1.dns-solutions.net       59m59s      
 2   shaharazran.com             NS    ns2.dns-solutions.net       59m59s      
 3   ns1.dns-solutions.net       A     69.12.120.3                 59m59s      
 4   ns2.dns-solutions.net       A     69.12.69.3                  59m59s      
[admin@mt] ip dns cache> 
[admin@mt] ip dns cache> /sys resou pr
                   uptime: 3d18h16m39s
                  version: "2.9.24"
              free-memory: 93592kB
             total-memory: 127700kB
                      cpu: "6x86"
                 cpu-load: 24
           free-hdd-space: 31640kB
          total-hdd-space: 60729kB
  write-sect-since-reboot: 5778
         write-sect-total: 621958
[admin@mt] ip dns cache>

I still think there is something wrong with the DNS cache… :cry: I’m going to generate a supout and send this one off to MT. I really hope this can get resolved. Again, I beleive this is a bug related to using the hotspot with dns.

6

Well, the watchdog timer kicked in this morning - router crashed and rebooted (a whopping 5 days uptime!). No supout was generated, even though this is enabled. :confused: