Ignore:
Timestamp:
Aug 7, 2009, 10:00:51 AM (13 years ago)
Author:
mitchb
Message:
Weight r-m the same as p-b and b-b because they're all uniprocessors
LVS shows that r-m is consistently handling far more connections
than p-b or b-b, as it should since we've been weighting them by
amount of RAM.  Unfortunately, the AMD SMP hangs have caused us to
make these VMs uniprocessors, and I suspect that r-m's repeated
brief SVN hiccups are due to handling more load than it's prepared
for.  So, until we can run SMPs again, number of virtual cores
trumps amount of RAM.
File:
1 edited

Legend:

Unmodified
Added
Removed
  • trunk/lvs/debian/config/etc/ha.d/ldirectord.cf

    r1268 r1292  
    1212        real=18.181.0.228:25 gate 6144
    1313        real=18.181.0.234:25 gate 3072
    14         real=18.181.0.235:25 gate 5120
     14        real=18.181.0.235:25 gate 3072
    1515        real=18.181.0.237:25 gate 3072
    1616        service=smtp
     
    2828        real=18.181.0.228 gate 6144
    2929        real=18.181.0.234 gate 3072
    30         real=18.181.0.235 gate 5120
     30        real=18.181.0.235 gate 3072
    3131        real=18.181.0.237 gate 3072
    3232        fallback=127.0.0.1 gate
Note: See TracChangeset for help on using the changeset viewer.