All Connections consistently dropped every 60 minutes

  • On a clean and simple install (single Wan, transparent NAT port 80, squid proxy ssl.), I am seeing a consistent drop of ALL active tcp connections every 60 minutes.  The specific time seems to be every 60 mins based on the last reboot time, ie, this is not a timed event based on an hourly event.  In the cron service, there are two "updatetable" functions that occur every */60 minutes to clear ssh lockouts and antivirus entries, but I tried disabling those to no avail.

    Googling revealed several posts indicating that there may be a tcp_Timeout in BSD of 3600 seconds, but the info seems sketchy.  I search sysctl net.inet and found only two entries that might align and those were net.inet.tcp.keepidle=7200000 and net.inet.tcp.hostcache.expire=3600.  I modified both of those, but there was no change in the drops.

    The problem is that there is a 20-30 second "outage" every hour and the users are getting annoyed by this.


  • Interesting - I have not been able to find any timers, but I did find that this problem does not impact any connections that are simply NAT'd, only connections via Squid.

    That led to this:

    My system is similar to the one in the article (Xeon Quad core with 4GB RAM and 100GB for caching). They say that this is likely a garbage collection and index rebuild issue.  I changed my cache high and low settings to be 1% apart. - waiting to see impact.

Log in to reply