[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [master] memory: decrease the GUI limits a bit.

On 05/14/2010 04:25 PM, Bill Nottingham wrote:

Ales Kozumplik (akozumpl redhat com) said:
diff --git a/isys/isys.h b/isys/isys.h
index e3cb1fc..811b51b 100644
--- a/isys/isys.h
+++ b/isys/isys.h
@@ -21,8 +21,9 @@
  #define H_ISYS

  #define MIN_RAM			262144 // 256 MB
-#define MIN_GUI_RAM		524288 // 512 MB
+#define GUI_INSTALL_EXTRA_RAM   131072 // 128 MB
  #define URL_INSTALL_EXTRA_RAM   131072 // 128 MB
  #define EARLY_SWAP_RAM		524288

What evidence is there that this is enough?


Good question! Simple answer: I tried and it is enough (more precisely: I tried a GUI installation, over nfs, with LVM+encryption, on a 384 MB virtual memory machine and got over the "Finding storage devices" step, that's the most critical, because GUI is already up and we make the LVM calls there)

More elaboratively: I had to increase the MIN_RAM couple weeks ago, see fbe24fe21350eec043d3f8ad41c1908b0878114a on master (there's also a similar patch for rhel6). The commit message shows some numbers too, but all of that was only tested during text install. It's very difficult to find out exactly how much X+metacity+gtk+other-desktop-crap will take together. I assumed I'd keep that at 256 MB extra, but then Hans complained he can install over http with GUI on 512 MB. So you will notice that while I increased the MIN_RAM by 128 MB I am decreasing the MIN_GUI_RAM by 128 MB now. This makes some sense: we are now running network manager (and who knows what else) in both text and gui install and that might use dynamic libraries that are later reused by the gui components (while earlier those were only loaded when X/gtk started).

That's my theory anyways but my experience seems to support it.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]