Bug 1284834

Introduced a new function called ctr_junkie to catch this special case were

!loc->name && loc->parent

(gdb) bt
#0  ctr_junkie (this=0x7f988000bb50, loc=0x7f9858001cbc) at changetimerecorder.c:315
#1  0x00007f98870e541b in ctr_lookup (frame=0x7f9870094eac, this=0x7f988000bb50, loc=0x7f9858001cbc, 
    xdata=0x7f985800099c) at changetimerecorder.c:334
#2  0x00007f9894c8d45e in default_lookup (frame=0x7f9870094eac, this=0x7f988000f360, loc=0x7f9858001cbc, 
    xdata=0x7f985800099c) at defaults.c:2377
#3  0x00007f98867d6efe in br_stub_lookup (frame=0x7f987009676c, this=0x7f9880011010, loc=0x7f9858001cbc, 
    xdata=0x7f985800099c) at bit-rot-stub.c:2709
#4  0x00007f98865b99b0 in posix_acl_lookup (frame=0x7f987009602c, this=0x7f98800126a0, loc=0x7f9858001cbc, 
    xattr=0x7f985800099c) at posix-acl.c:909
#5  0x00007f98863a6ea9 in pl_lookup (frame=0x7f9870000bac, this=0x7f9880013b30, loc=0x7f9858001cbc, 
    xdata=0x7f985800099c) at posix.c:2171
#6  0x00007f988618cba2 in up_lookup (frame=0x7f987001444c, this=0x7f9880014f60, loc=0x7f9858001cbc, 
    xattr_req=0x7f985800099c) at upcall.c:775
#7  0x00007f9894c892b7 in default_lookup_resume (frame=0x7f9858000f4c, this=0x7f9880016570, loc=0x7f9858001cbc, 
    xdata=0x7f985800099c) at defaults.c:1731
#8  0x00007f9894c1d18d in call_resume_wind (stub=0x7f9858001c7c) at call-stub.c:2223
#9  0x00007f9894c2504e in call_resume (stub=0x7f9858001c7c) at call-stub.c:2571
#10 0x00007f9885f7f880 in iot_worker (data=0x7f98800445c0) at io-threads.c:210
#11 0x00007f9893a5452a in start_thread () from /lib64/libpthread.so.0
#12 0x00007f98933a322d in clone () from /lib64/libc.so.6

there was another function inserted in marker_lookup called marker_junkie.
This was done to check if this lookup came from server xlator or marker translator.

This is how the code looks in marker

This is how the test code looks in ctr

void ctr_junkie (xlator_t *this, loc_t *loc) {
}

int32_t
ctr_lookup (call_frame_t *frame, xlator_t *this,
              loc_t *loc, dict_t *xdata)
{
        gf_ctr_inode_context_t ctr_inode_cx;
        gf_ctr_inode_context_t *_inode_cx       = &ctr_inode_cx;
        gf_ctr_link_context_t  ctr_link_cx;
        gf_ctr_link_context_t  *_link_cx        = &ctr_link_cx;
        int ret                                 = -1;

        CTR_IS_DISABLED_THEN_GOTO(this, out);
        CTR_IF_INTERNAL_FOP_THEN_GOTO (frame, xdata, out);

        GF_ASSERT(frame);
        GF_ASSERT(frame->root);

        if (loc->parent && !loc->name) {
                ctr_junkie (this, loc); ------------> catch the special case and call ctr_junkie
        }



And this is how the test code looks in marker 
void
marker_junkie (xlator_t *this, loc_t *loc) {
}

int32_t
marker_lookup (call_frame_t *frame, xlator_t *this,
               loc_t *loc, dict_t *xattr_req)
{
        int32_t         ret                     = 0;
        marker_local_t *local                   = NULL;
        marker_conf_t  *priv                    = NULL;

        priv = this->private;

        if (loc->parent && !loc->name) {
                marker_junkie (this, loc);------------> catch the special case and call marker_junkie
        }

The effect of this is the ctr lookup heal that is triggered for legacy files is not done. Functionality break. The databases are empty.

Plus looks like vanila quota is broken i,.e with tiering on a normal volume


[root@fedora1 test]# gluster vol quota test list
                  Path                   Hard-limit  Soft-limit      Used  Available  Soft-limit exceeded? Hard-limit exceeded?
-------------------------------------------------------------------------------------------------------------------------------
/                                        20Bytes     80%(16Bytes)    1.5MB  0Bytes             Yes                  Yes


I still can create files

(In reply to Joseph Elwin Fernandes from comment #7)
> Plus looks like vanila quota is broken i,.e with tiering on a normal volume
> 
> 
> [root@fedora1 test]# gluster vol quota test list
>                   Path                   Hard-limit  Soft-limit      Used 
> Available  Soft-limit exceeded? Hard-limit exceeded?
> -----------------------------------------------------------------------------
> --------------------------------------------------
> /                                        20Bytes     80%(16Bytes)    1.5MB 
> 0Bytes             Yes                  Yes
> 
> 
> I still can create files

typo without tiering on a normal volume

The logs are getting generated due to a nameless lookup. But couldn't see any kind of loss in data even after attach-tier. So it's better to avoid nameless lookup in ctr, thereby avoiding the log.

https://code.engineering.redhat.com/gerrit/#/c/63434/

After attach tier , not seeing any error logs on latest build. 

Bug verified on build glusterfs-3.7.5-11.el7rhgs.x86_64.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://rhn.redhat.com/errata/RHBA-2016-0193.html