Is there a way to disable "fastest mirror search"

[Emmanuel Charpentier]

Dear all,

The new all-shiny all-singing all-dancing fastest mirror selection is probably useful for large tasks (maybe not so much for small spkgs). I note however that it is launced once fopr each package/tarball/item that has to be downloaded, not once per job.

Furthermore, behind some firewalls (most notably those that block ICMP), this feature is a royal pain in the ass^Hymptote : mirror selection failure, hence task failure, hence job failure. Hosed upgrade.

I foud no way to disable this err...feature  Defining SAGE_SERVER is ignored.

Am I missearching the doc ? Any other idea ?

--
Emmanuel Charpentier

[Volker Braun]

The timings are cached (assuming that they find a mirror).

The timing is based on an actual connection on port 80, no ICMP involved I think. If your firewall blocks it then it will also block downloads.

[Emmanuel Charpentier]

Our firewall attempts to filter downloads (for virus scanning purpose, or se we are told) more or less transparently : when I use wget, I see a waitong period with on byte per second, then a fast (Ethernet speed)) download of the target file.

IIRC, you evoked the possibility of ICMP-induced trouble during a previous episode of firewall-related trouble. What did you have in mind ?

Anyway : any idea to disable this mirror selection ?

--
Emmanuel Charpentier

[Volker Braun]

On Tuesday, June 16, 2015 at 3:37:30 PM UTC+2, Emmanuel Charpentier wrote:

IIRC, you evoked the possibility of ICMP-induced trouble during a previous episode of firewall-related trouble. What did you have in mind ?

I had just forgotten what I did; the fastest mirror code does not use ICMP. 

Can you post the actual output?

[Emmanuel Charpentier]

I'll have a look tomorrow (back at home...).

--
Emmanuel Charpentier
 

[Emmanuel Charpentier]

Le mardi 16 juin 2015 19:03:27 UTC+2, Volker Braun a écrit :

On Tuesday, June 16, 2015 at 3:37:30 PM UTC+2, Emmanuel Charpentier wrote:

IIRC, you evoked the possibility of ICMP-induced trouble during a previous episode of firewall-related trouble. What did you have in mind ?

I had just forgotten what I did;

In this thread,  tenth post from the top (sorry : I use Google groups, which doe not give very precise references...)

the fastest mirror code does not use ICMP. 

HTH,
--
Emmanuel Charpentier

[Volker Braun]

On Tuesday, June 16, 2015 at 9:43:21 PM UTC+2, Emmanuel Charpentier wrote:

In this thread,  tenth post from the top (sorry : I use Google groups, which doe not give very precise references...)

the fastest mirror code does not use ICMP. 

Yes and that is wrong. I had forgotten what I wrote because it was many months ago (=review duration). 

Post your logs

[Emmanuel Charpentier]

Okay. I went to my $SAGE_ROOT. Compressed my install.log (about 20 Mb native...).

I tried to illustrate my point by attempting to install gap_packages. First trial : ./sage-i gap_packages. Results in gap_packages-T1.log.

Second trial : SAGE_SERVER=http://www-ftp.lip6.fr/pub/math/sagemath/ ./sage -i gap_packages. Result in gap_packages-T2.log (identical as far as I can tell).

The only way I found to install a package was to download manually the right tarball in $SAGE_ROOT/upstream and to install it via ./sage -i <name of the package without version>

HTH,
--
Emmanuel Charpentier

[Volker Braun]

delete SAGE_ROOT/upstream/mirror_list and try again

[Emmanuel Charpentier]

Le mercredi 17 juin 2015 12:41:17 UTC+2, Volker Braun a écrit :

delete SAGE_ROOT/upstream/mirror_list and try again

Nothing even remotely like that in my installation :

charpent@SAP5057241:/usr/local/sage-6.7$ find . -iname "*mirror*list*"
charpent@SAP5057241:/usr/local/sage-6.7$

[Volker Braun]

Should be fixed in http://trac.sagemath.org/ticket/18723, can you give it a try?

[Emmanuel Charpentier]

Mixed results : OK for installing a package (sage -i ...), not for enquiring (sage -optional). See ticket for details.