Problem with Solaris

Server NGINX
1

Hello,

I just came to this old problem :
readv() failed (22: Invalid argument) while reading upstream
Running nginx 0.8.52 on Solaris10

I have seen a message from 2008 that spoke about that, but it seems that
the bug is not yet resolved.
Or is there any workaround ?

Thanks in advance…

f.g.

–
Frdric Goudal
Service Informatique
ENSEIRB-MATMECA / IPB
Tel : +33 556 84 23 11
Fax: +33 556 37 20 23

2

Hello!

On Tue, Oct 26, 2010 at 03:55:38PM +0200, Frédéric Goudal wrote:

Hello,

I just came to this old problem :
readv() failed (22: Invalid argument) while reading upstream
Running nginx 0.8.52 on Solaris10

I have seen a message from 2008 that spoke about that, but it seems that the bug
is not yet resolved.
Or is there any workaround ?

The only thread I see in archives is this one (in Russian):

There Igor explains that there are two possible causes:

  1. (An unknown) bug in nginx.

  2. Known problem in Solaris which sets errno to EINVAL (instead of
    ECONNRESET as others do and POSIX requires) when connection is
    reset by peer (i.e. your backend).

I believe (2) is much more probable (read: check your backend).

If you think that cause is (1) you may want to provide more
debugging info, see here:

Maxim D.

3

I agree, I have found the same message.

The disturbing thing is that, while trying to dig a bit more, the last
time I launched nginx, the bug did not appear.

I know that Solaris is not a trendy platform… but… I have no other
solution.

It’s strange that nobody else use that OS :slight_smile:

Well… I guess I have to dig by myself… or is there anybody wanting
to help ? Even in trying to correct the bug ?

f.g.

Le 26 oct. 2010 21:16, Maxim D. a crit :

I have seen a message from 2008 that spoke about that, but it seems that the
bug is not yet resolved.

  1. Known problem in Solaris which sets errno to EINVAL (instead of
    Maxim D.

nginx mailing list
[email protected]
mailman.nginx.org archive

–
Frdric Goudal
Service Informatique
ENSEIRB-MATMECA / IPB
Tel : +33 556 84 23 11
Fax: +33 556 37 20 23

4

Le 26 oct. 2010 22:23, Bryan Horstmann-Allen a crit :

|
| Well… I guess I have to dig by myself… or is there anybody wanting to
help ? Even in trying to correct the bug ?

I use nginx on Solaris 10 and have no had this problem, that I recall. I use
the NetBSD pkgsrc ports tree; you might want to see their patches directory to
see if there’s anything relevant.

The only thing off the top of my head I can think of… have you set “sendfile
on;”?

I had it on first than removed it on second as I saw it could cause a
problem.
But than… it seemed I still had the problem.
Ok I started over and I did not found the problem again. So maybe an
error from me ?

Do you know why sendfile on can cause the problem ?

f.g.

–
Frdric Goudal
Service Informatique
ENSEIRB-MATMECA / IPB
Tel : +33 556 84 23 11
Fax: +33 556 37 20 23

5

±-----------------------------------------------------------------------------
| On 2010-10-26 22:04:56, Fr?d?ric Goudal wrote:
|
| I agree, I have found the same message.
|
| The disturbing thing is that, while trying to dig a bit more, the last time I
launched nginx, the bug did not appear.
|
| I know that Solaris is not a trendy platform… but… I have no other solution.
|
| It’s strange that nobody else use that OS :slight_smile:
|
| Well… I guess I have to dig by myself… or is there anybody wanting to help
? Even in trying to correct the bug ?

I use nginx on Solaris 10 and have no had this problem, that I recall. I
use
the NetBSD pkgsrc ports tree; you might want to see their patches
directory to
see if there’s anything relevant.

The only thing off the top of my head I can think of… have you set
“sendfile
on;”?

bdha
cyberpunk is dead. long live cyberpunk.