[Labs-l] Out of memory errors

Mon Aug 3 16:43:30 UTC 2015

I wrote:

> [...]

> The basic cause is that the available virtual memory is
> overstated by these hosts as the jobs running there will
> share substantial amounts of memory by using the same bina-
> ries (lighttpd, php-cgi, etc.).  If one of those web ser-
> vices does something different, then the formula doesn't
> work anymore and the host runs short on real memory.

> [...]

A little more digging: No :-) (I think).

For normal execution nodes, we set a complex_value h_vmem;
but the webgrid nodes have none.  So I assume the grid just
distributes the web services by load.

I'll calculate the current (theoretical) virtual memory con-
sumption of the webgrid nodes based on the jobs running on
them and then set a value ad interim that makes sense to me
(cf. also https://phabricator.wikimedia.org/T107665 ("Deter-
mine and deploy proper h_vmem resources for execution
nodes")).

Tim