[infiniband-diags] [0/3] support --diff and --diffcheck in ibnetdiscover

[infiniband-diags] [0/3] support --diff and --diffcheck in ibnetdiscover

[Al Chu]

Hey Sasha,

The following sets of patches implement a --diff and --diffcheck options
in ibnetdiscover to let users diff an ibnetdiscover state to a previous
ibnetdiscover state.  The goal of this option is to help system
administrators isolate/determine changes in the network quickly compared
to a previous state.  Here's an example:

# > ./ibnetdiscover --diff=orig.cache 

vendid=0x8f1
devid=0x5a30
sysimgguid=0x8f10400411f57
switchguid=0x8f10400411f56(8f10400411f56)
Switch  24 "S-0008f10400411f56"         # "ISR9024D Voltaire" base port 0 lid 11 lmc 0
< [14]  "H-0002c90200219ef0"[1](2c90200219ef1)          # "wopr0" lid 64 4xDDR
< [19]  "H-0002c9030000ff7c"[1](2c9030000ff7d)          # "wopr9" lid 48 4xDDR
> [20]  "H-0002c9030000ff7c"[1](2c9030000ff7d)          # "wopr9" lid 4 4xDDR

< vendid=0x2c9
< devid=0x6282
< sysimgguid=0x2c90200219ef3
< caguid=0x2c90200219ef0
< Ca    2 "H-0002c90200219ef0"          # "wopr0"
< [1](2c90200219ef1)    "S-0008f10400411f56"[14]                # lid 64 lmc 2 "ISR9024D Voltaire" lid 11 4xDDR

In this particular example, port 14 on the switch (which is connected to
node 'wopr0') was up before but is now down (and the associated CA is
noted too).  In addition, 'wopr9' is connected to port 20 instead of
port 19 on the switch.

By default --diff checks switches, cas, routers, and port connections.
The --diffcheck option allows the user to specify which diff options
they want done, and also adds other diff checks for lids and/or node
descriptions.  More diff checks could be added later as needed.  For
example, the following only checks for differences of lids on switches.

# > ./ibnetdiscover --diff=orig.cache --diffcheck=sw,lid

vendid=0x8f1
devid=0x5a30
sysimgguid=0x8f10400411f57
switchguid=0x8f10400411f56(8f10400411f56)
< Switch        24 "S-0008f10400411f56"         # "ISR9024D Voltaire" base port 0 lid 11 lmc 0
> Switch        24 "S-0008f10400411f56"         # "ISR9024D Voltaire" base port 0 lid 3 lmc 0
< [13]  "H-0002c90200219e64"[1](2c90200219e65)          # "wopri" lid 4 4xDDR
> [13]  "H-0002c90200219e64"[1](2c90200219e65)          # "wopri" lid 1 4xDDR

Others on the list may wonder how this is different than just using the
normal 'diff' tool.  The differences I can think of are:

1) This checks differences in the network, not text.  This is
particularly important when lids, lmc, etc. are changed.  Otherwise
there are many differences in a normal diff output that aren't
necessary.

2) This provides the appropriate "context" in the diff output, showing
the appropriate system ids to allow a system administrator to identify
ports on what switch have changed.  Under normal diff output, you may
not get that appropriate context of information.  The system
administrator can of course use options like --context in diff, but the
goal is to make the diff output clear and concise, not outputting
unnecessary junk.

3) As parallelization has been added into ibnetdisocver/libibnetdiscover
this becomes more critical as output in ibnetdiscover/libibnetdiscover
can be re-ordered.  So a normal diff suddenly is non-functional.

There's probably other minor advantages.  Even if minor output tweaks
happen to ibnetdiscover in the future, this can still work against old
cache files.

Al


-- 
Albert Chu
chu11-i2BcT+NCU+M@public.gmane.org
Computer Scientist
High Performance Systems Division
Lawrence Livermore National Laboratory

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [infiniband-diags] [0/3] support --diff and --diffcheck in ibnetdiscover

[Sasha Khapyorsky]

Hi Al,

On 10:05 Wed 07 Apr     , Al Chu wrote:
> 
> Others on the list may wonder how this is different than just using the
> normal 'diff' tool.  The differences I can think of are:
> 
> 1) This checks differences in the network, not text.  This is
> particularly important when lids, lmc, etc. are changed.  Otherwise
> there are many differences in a normal diff output that aren't
> necessary.
> 
> 2) This provides the appropriate "context" in the diff output, showing
> the appropriate system ids to allow a system administrator to identify
> ports on what switch have changed.  Under normal diff output, you may
> not get that appropriate context of information.  The system
> administrator can of course use options like --context in diff, but the
> goal is to make the diff output clear and concise, not outputting
> unnecessary junk.
> 
> 3) As parallelization has been added into ibnetdisocver/libibnetdiscover
> this becomes more critical as output in ibnetdiscover/libibnetdiscover
> can be re-ordered.  So a normal diff suddenly is non-functional.

I'm getting your arguments. And this remind me the question which was
already raised some time ago.

Would it be better to keep cache in a regular human readable
ibnetdiscover output format, so '--diff' will be usable not just against
cache, but also against a regular ibnetdiscover output files?

Sasha
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [infiniband-diags] [0/3] support --diff and --diffcheck in ibnetdiscover

[Sasha Khapyorsky]

On 16:08 Tue 13 Apr     , Sasha Khapyorsky wrote:
> 
> Would it be better to keep cache in a regular human readable
> ibnetdiscover output format, so '--diff' will be usable not just against
> cache, but also against a regular ibnetdiscover output files?

BTW, such implementation could be simplified by the fact that we have
already ibnetdiscover output parser in 'ibsim'.

Sasha
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [infiniband-diags] [0/3] support --diff and --diffcheck in ibnetdiscover

[Al Chu]

Hey Sasha,

On Tue, 2010-04-13 at 06:10 -0700, Sasha Khapyorsky wrote:
> On 16:08 Tue 13 Apr     , Sasha Khapyorsky wrote:
> > 
> > Would it be better to keep cache in a regular human readable
> > ibnetdiscover output format, so '--diff' will be usable not just against
> > cache, but also against a regular ibnetdiscover output files?

I had considered this at one point.  There were several reasons I
decided to go w/ the cache idea.  Perhaps the major reason is that the
current cache system has "all" the data (nodeinfo, portinfo, etc.)
saved, whereas the normal ibnetdiscover output does not.  So we would be
limited in our diff output for what ibnetdiscover outputs (or perhaps
limited in future extensions).  Adding a --diff into iblinkinfo is also
on my list to do, but that wouldn't be possible w/ a cached text output
(the cached text file doesn't store portinfo data).

This isn't to say there are downsides.  The major downside is that it's
difficult to edit the cache when minor changes happen on the network
(e.g. for our system administrators, when an HCA is replaced b/c a node
dies).  I have a tool for that (to be submitted soon too I hope :-).

Al

-- 
Albert Chu
chu11-i2BcT+NCU+M@public.gmane.org
Computer Scientist
High Performance Systems Division
Lawrence Livermore National Laboratory

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [infiniband-diags] [0/3] support --diff and --diffcheck in ibnetdiscover

[Sasha Khapyorsky]

On 10:17 Tue 13 Apr     , Al Chu wrote:
> 
> I had considered this at one point.  There were several reasons I
> decided to go w/ the cache idea.  Perhaps the major reason is that the
> current cache system has "all" the data (nodeinfo, portinfo, etc.)
> saved, whereas the normal ibnetdiscover output does not. So we would be
> limited in our diff output for what ibnetdiscover outputs (or perhaps
> limited in future extensions).  Adding a --diff into iblinkinfo is also
> on my list to do, but that wouldn't be possible w/ a cached text output
> (the cached text file doesn't store portinfo data).

That makes sense.

> This isn't to say there are downsides.  The major downside is that it's
> difficult to edit the cache when minor changes happen on the network
> (e.g. for our system administrators, when an HCA is replaced b/c a node
> dies). I have a tool for that (to be submitted soon too I hope :-).

It is difficult to edit and difficult to read. So I would really prefer
to have a text file (even with extended format). But well, let's see how
this will go.

Sasha
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html