Re: RGW won't start after upgrade to 12.2.5

[Marc Spencer <mspencer@xxxxxxxxxxxxxxxx>]

This is now filed as bug #24228

 

Marc D. Spencer Chief Technology Officer T: 866.808.4937 × 202 E: mspencer@xxxxxxxxxxxxxxxx www.liquidpixels.com

On May 21, 2018, at 10:58 PM, Marc Spencer <mspencer@xxxxxxxxxxxxxxxx> wrote:

I found the issue, for the curious.

The default configuration for rgw_ldap_secret seems to be set to /etc/openldap/secret, which on my system is empty:

# ceph-conf -D | grep ldap
rgw_ldap_binddn = uid=admin,cn=users,dc=example,dc=com
rgw_ldap_dnattr = uid
rgw_ldap_searchdn = cn=users,cn=accounts,dc=example,dc=com
rgw_ldap_searchfilter = 
rgw_ldap_secret = /etc/openldap/secret
rgw_ldap_uri = ldaps://<ldap.your.domain>
rgw_s3_auth_use_ldap = false

# cat /etc/openldap/secret
cat: /etc/openldap/secret: No such file or directory

But the code assumes that if it is set, the named file has content. Since it doesn’t, safe_read_file() asserts.

I set it to nothing (rgw_ldap_secret = ) in my configuration, and everything seems happy.

std::string parse_rgw_ldap_bindpw(CephContext* ctx)

{
  string ldap_bindpw;
  string ldap_secret = ctx->_conf->rgw_ldap_secret;

  if (ldap_secret.empty()) {
    ldout(ctx, 10)
      << __func__ << " LDAP auth no rgw_ldap_secret file found in conf"
      << dendl;
    } else {
      char bindpw[1024];
      memset(bindpw, 0, 1024);
      int pwlen = safe_read_file("" /* base */, ldap_secret.c_str(),
                                 bindpw, 1023);
    if (pwlen) {
      ldap_bindpw = bindpw;
      boost::algorithm::trim(ldap_bindpw);
      if (ldap_bindpw.back() == '\n')
        ldap_bindpw.pop_back();
    }
  }

  return ldap_bindpw;
}

On May 21, 2018, at 5:27 PM, Marc Spencer <mspencer@xxxxxxxxxxxxxxxx> wrote:

Hi,

  I have a test cluster of 4 servers running Luminous. We were running 12.2.2 under Fedora 17 and have just completed upgrading to 12.2.5 under Fedora 18.

  All seems well: all MONs are up, OSDs are up, I can see objects stored as expected with rados -p default.rgw.buckets.data ls. 

  But when i start RGW, my load goes through the roof as radosgw continuously rapid-fire core dumps. 

Log Excerpt:

… 

   -16> 2018-05-21 15:52:48.244579 7fc70eeda700  5 -- 10.19.33.13:0/3446208184 >> 10.19.33.14:6800/1417 conn(0x55e78a610800 :-1 s=STATE_OPEN_MESSAGE_READ_FOOTER_AND_DISPATCH pgs=14567 cs=1 l=1). rx osd.6 seq 7 0x55e78a67b500 osd_op_reply(47 notify.6 [watch watch cookie 94452947886080] v1092'43446 uv43445 _ondisk_ = 0) v8
   -15> 2018-05-21 15:52:48.244619 7fc70eeda700  1 -- 10.19.33.13:0/3446208184 <== osd.6 10.19.33.14:6800/1417 7 ==== osd_op_reply(47 notify.6 [watch watch cookie 94452947886080] v1092'43446 uv43445 _ondisk_ = 0) v8 ==== 152+0+0 (1199963694 0 0) 0x55e78a67b500 con 0x55e78a610800
   -14> 2018-05-21 15:52:48.244777 7fc723656000  1 -- 10.19.33.13:0/3446208184 --> 10.19.33.15:6800/1433 -- osd_op(unknown.0.0:48 16.1 16:93e5b521:::notify.7:head [create] snapc 0=[] ondisk+write+known_if_redirected e1092) v8 -- 0x55e78a67bc00 con 0
   -13> 2018-05-21 15:52:48.275650 7fc70eeda700  5 -- 10.19.33.13:0/3446208184 >> 10.19.33.15:6800/1433 conn(0x55e78a65e000 :-1 s=STATE_OPEN_MESSAGE_READ_FOOTER_AND_DISPATCH pgs=14572 cs=1 l=1). rx osd.2 seq 7 0x55e78a678380 osd_op_reply(48 notify.7 [create] v1092'43453 uv43453 _ondisk_ = 0) v8
   -12> 2018-05-21 15:52:48.275675 7fc70eeda700  1 -- 10.19.33.13:0/3446208184 <== osd.2 10.19.33.15:6800/1433 7 ==== osd_op_reply(48 notify.7 [create] v1092'43453 uv43453 _ondisk_ = 0) v8 ==== 152+0+0 (2720997170 0 0) 0x55e78a678380 con 0x55e78a65e000
   -11> 2018-05-21 15:52:48.275849 7fc723656000  1 -- 10.19.33.13:0/3446208184 --> 10.19.33.15:6800/1433 -- osd_op(unknown.0.0:49 16.1 16:93e5b521:::notify.7:head [watch watch cookie 94452947887232] snapc 0=[] ondisk+write+known_if_redirected e1092) v8 -- 0x55e78a688000 con 0
   -10> 2018-05-21 15:52:48.296799 7fc70eeda700  5 -- 10.19.33.13:0/3446208184 >> 10.19.33.15:6800/1433 conn(0x55e78a65e000 :-1 s=STATE_OPEN_MESSAGE_READ_FOOTER_AND_DISPATCH pgs=14572 cs=1 l=1). rx osd.2 seq 8 0x55e78a688000 osd_op_reply(49 notify.7 [watch watch cookie 94452947887232] v1092'43454 uv43453 _ondisk_ = 0) v8
    -9> 2018-05-21 15:52:48.296824 7fc70eeda700  1 -- 10.19.33.13:0/3446208184 <== osd.2 10.19.33.15:6800/1433 8 ==== osd_op_reply(49 notify.7 [watch watch cookie 94452947887232] v1092'43454 uv43453 _ondisk_ = 0) v8 ==== 152+0+0 (3812136207 0 0) 0x55e78a688000 con 0x55e78a65e000
    -8> 2018-05-21 15:52:48.296924 7fc723656000  2 all 8 watchers are set, enabling cache
    -7> 2018-05-21 15:52:48.297135 7fc57cbb6700  2 garbage collection: start
    -6> 2018-05-21 15:52:48.297185 7fc57c3b5700  2 object expiration: start
    -5> 2018-05-21 15:52:48.297321 7fc57cbb6700  1 -- 10.19.33.13:0/3446208184 --> 10.19.33.16:6804/1596 -- osd_op(unknown.0.0:50 18.3 18:d242335b:gc::gc.2:head [call lock.lock] snapc 0=[] ondisk+write+known_if_redirected e1092) v8 -- 0x55e78a692000 con 0
    -4> 2018-05-21 15:52:48.297395 7fc57c3b5700  1 -- 10.19.33.13:0/3446208184 --> 10.19.33.16:6804/1596 -- osd_op(unknown.0.0:51 18.0 18:1a734c59:::obj_delete_at_hint.0000000000:head [call lock.lock] snapc 0=[] ondisk+write+known_if_redirected e1092) v8 -- 0x55e78a692380 con 0
    -3> 2018-05-21 15:52:48.299463 7fc568b8e700  5 schedule life cycle next start time: Tue May 22 04:00:00 2018
    -2> 2018-05-21 15:52:48.299528 7fc567b8c700  5 ERROR: sync_all_users() returned ret=-2
    -1> 2018-05-21 15:52:48.299698 7fc56738b700  1 -- 10.19.33.13:0/3446208184 --> 10.19.33.14:6800/1417 -- osd_op(unknown.0.0:52 18.7 18:e9187ab8:reshard::reshard.0000000000:head [call lock.lock] snapc 0=[] ondisk+write+known_if_redirected e1092) v8 -- 0x55e78a54fc00 con 0
     0> 2018-05-21 15:52:48.301978 7fc723656000 -1 *** Caught signal (Aborted) **
 in thread 7fc723656000 thread_name:radosgw

 ceph version 12.2.5 (cad919881333ac92274171586c827e01f554a70a) luminous (stable)
 1: (()+0x22d82c) [0x55e7861a882c]
 2: (()+0x11fb0) [0x7fc719270fb0]
 3: (gsignal()+0x10b) [0x7fc716603f4b]
 4: (abort()+0x12b) [0x7fc7165ee591]
 5: (parse_rgw_ldap_bindpw[abi:cxx11](CephContext*)+0x68b) [0x55e78647409b]
 6: (rgw::auth::s3::LDAPEngine::init(CephContext*)+0xb9) [0x55e7863a38f9]
 7: (rgw::auth::s3::ExternalAuthStrategy::ExternalAuthStrategy(CephContext*, RGWRados*, rgw::auth::s3::AWSEngine::VersionAbstractor*)+0x74) [0x55e786154bc4]
 8: (std::__shared_ptr<rgw::auth::StrategyRegistry, (__gnu_cxx::_Lock_policy)2>::__shared_ptr<std::allocator<rgw::auth::StrategyRegistry>, CephContext* const&, RGWRados* const&>(std::_Sp_make_shared_tag, std::allocator<rgw::auth::StrategyRegistry> const&, CephContext* const&, RGWRados* const&)+0xf8) [0x55e786158f78]
 9: (main()+0x196b) [0x55e78614463b]
 10: (__libc_start_main()+0xeb) [0x7fc7165f01bb]
 11: (_start()+0x2a) [0x55e78614c3da]
 NOTE: a copy of the executable, or `objdump -rdS <executable>` is needed to interpret this.

 

Marc D. Spencer Chief Technology Officer T: 866.808.4937 × 202 E: mspencer@xxxxxxxxxxxxxxxx www.liquidpixels.com

_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com

 

Marc D. Spencer Chief Technology Officer T: 866.808.4937 × 202 E: mspencer@xxxxxxxxxxxxxxxx www.liquidpixels.com

_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com