Thursday, November 25, 2010

ghetto lustre bug nearly donates my whole group a set of fresh pink slips

To: Prof. of famous discovery; Prof. of very famous things;
Cc: Some Assorted Deans

From: James Cuff
Subject: Groveling 

Dear Professors of all career holding...

We are still inordinately sorry about that whole missing set of galaxies thing, and those few lost irreplaceable genomes that kinda happened as a result of our scratch disks conking out. Although I do have a wizard idea where we can most likely write this all up together in Science as a new complex black hole / mythical dark matter or junk dna phenomenon? Yeah, I know maybe not; you each always told me that we best leave such rabid extrapolations to our atmospheric chemistry department...

Anyway, I digress. I've attached our latest in-depth analysis from our best top industry experts below. I hope this is enough information to not get my whole team fired?

To: one of my most epic team members
From: a very smart Lustre developer that totally saved our bacon (I hope)!
Subject: [Lustre-discuss] non-consecutive OST ordering

This is a bug in llapi_lov_get_uuids() which assigns UUID to the wrong OST index when there are sparse OST(s).  Please file a bug for this. Before this bug can be fixed, you can apply the following patch to e2fsprogs(version lfsck.c as a workaround (not verified though).
e2fsprogs/e2fsck/lfsck.c   2010-11-12 11:43:42.000000000 +0800
+++ lfsck.c
2010-11-12 12:14:38.000000000 +0800
@@ -1226,6 +1226,12 @@
    int i, rc;
 +   /* skip empty UUID OST */
 if(!strlen(lfsck_uuid[ost_idx].uuid)) {
 +       log_write("index %d UUID is
 empty(sparse OST index?). Skipping.\n", ost_idx);
 +       return(0);

p.s. for those w/o a sense of humor this one is all totally fictional - well parts of it certainly are at any rate ;-)

Remember you may pick from only two things, either fast or cheap.  However, you may pick only but one!

A very happy start of holidays to all!

(c) 2018 James Cuff