hellomynameisjoel's questions

I am creating a stack via the AWS Ruby SDK v3 and it is failing with an internal error. After a long period of time (30-40 minutes) it fails and rolls back with an internal failure. No resources show up and there are no other events.

24 May 2019 07:32:08    stack-name ROLLBACK_COMPLETE    -
24 May 2019 07:31:42    stack-name ROLLBACK_IN_PROGRESS Internal Failure. Rollback requested by user.
24 May 2019 07:05:10    stack-name CREATE_IN_PROGRESS   User Initiated

Launching the stack as a different (more privileged) user means that it starts immediately. I have also tried removing all resources from the stack except for one EC2 instance and the same behaviour is exhibited.

The stack is launched with capabilities CAPABILITY_IAM and CAPABILITY_NAMED_IAM . My best guess is that the user is missing an IAM policy which allows it to even start processing the stack, but I don't know how to debug it. It's a role that is used by many other deployments and cannot be easily changed. Any guidance on how to debug or what might cause this behaviour is appreciated.

The code does produce a large list of dynamically produced Parameters and Tags. From what I can tell these are working correctly as they show up in the console as expected.

I have a requirement to give a user access to an NFS share on which they may place some files, and then delete the files. Assume they are rogue and may try to really delete the files (I'm not sure if shred works via NFS but whatever is the equivalent). We don't want them to realise the NFS share has been changed (it's their home directory).

Yeah... strange requirement I know!

Is there a filesystem or setting on NFS so it appears like the user is deleting files while really they are being stored? I was told this was a revisioning file system but I could not find many references to that.

I'm aware you can "undelete" on ext3 and similar by using recovery programs but we would like something more robust than that, ideally something that stores all copies of files in some easily accessible form.

Using CentOS 5.8 . User can't delete their files because they are "rogue" and no longer trusted (it's complicated, and that's all I know).

We have a code, content and DB flow that follows a typical pattern:

Dev->Build->Staging->Prod

The kicker is that the client is editing and publishing content only in Prod.

So our current strategy is to attempt to do a full content sync Prod->Staging on a semi-regular basis and then a full sync Prod->Build/Dev on an ad-hoc basis. Doing a Prod->Build/Dev push means the environment is unusable for a period of time (at least a day).

What strategies exist to make sure your dev/test/build environments accurately match prod in terms of content and how are others managing this problem?

We are using Sitecore 6.1 as the CMS system

The situation is that we have a number of guests, say 80, sending through a central mail infrastructure out to the web. The way the network is setup the mail server cannot necessarily initiate connections back to each of the senders.

I understand I can disable bounce messages completely by turning disabling the bounce daemon but I'm concerned if I do this I will miss system-problems. What would be good if all bounces went to postmaster and I could look for them there. There is no reason to ever try to contact the senders back.

Any ideas how to achieve this? Googling mentioned NOTIFY=NEVER and some stuff about smtpd_command_filter although what I'm doing feels like it should be simpler than that. I can get the bounce messages to postmaster using notify_classes but I think that relies upon the bounce daemon.

I'm running CentOS 5.5 and postfix 2.3.3-2.1.el5_2

EDIT:

soft_bounce=yes seems to be another way to disable bounces, but this is not exactly what I want

We wish to store some data in 389 Directory Server LDAP that doesn't fit that well into the standard set of schema's that come with the product. Nothing too amazing, things like:

• when the customer joined
• are they currently active
• customer certificate[1]
• which environment they are using

My question is this: should we register with OID and start writing up our own custom schema OR is there a standard schema definition not provided by Directory Server that we can download and use that would fit our needs? Should we munge/hack existing attributes and store the data among there (I'm strongly opposed to this, but would be interested in arguments about why its better than extending)?

[1] I know there is a field for this userCertificate but we don't want to use it to authenticate the user for the purposes of binding

Using CentOS 5.5 with 389 Directory Server 8.1

I have a DRBD partition mounted get auto-mounted using Linux-HA, then it sets up a bind and shares the directory via NFS. This shares out a directory for user home directories.

Some users work fine, most do not and it sits there hanging on mounting their folder. When I do a tcpdump I see:

07:45:16.415649 IP foo.server.com.1002 > nfs.server.com.nfs: . ack 976 win 348
07:45:22.705531 IP foo.server.c.2174542926 > nfs.server.com.nfs: 192 getattr [|nfs]
07:45:22.705643 IP nfs.server.com.nfs > foo.server.com.1002: . ack 1985 win 501
07:45:23.703517 IP nfs.server.com.nfs > foo.server.c.2174542926: reply ok 96 getattr ERROR: Request couldn't be completed in time
07:45:23.703564 IP foo.server.com.1002 > nfs.server.com.nfs: . ack 1072 win 348
07:45:25.053519 IP foo.server.c.2191320142 > nfs.server.com.nfs: 192 getattr [|nfs]
07:45:25.053651 IP nfs.server.com.nfs > foo.server.com.1002: . ack 2177 win 501
07:45:26.051473 IP nfs.server.com.nfs > foo.server.c.2191320142: reply ok 96 getattr ERROR: Request couldn't be completed in time
07:45:26.051522 IP foo.server.com.1002 > nfs.server.com.nfs: . ack 1168 win 348
07:45:31.413431 IP foo.server.c.2208097358 > nfs.server.com.nfs: 160 getattr [|nfs]
07:45:31.413556 IP nfs.server.com.nfs > foo.server.com.1002: . ack 2337 win 501
07:45:32.411393 IP nfs.server.com.nfs > foo.server.c.2208097358: reply ok 76 getattr ERROR: Request couldn't be completed in time

my auto.master looks like:

/home/users   /etc/auto.home  --timeout=60

my auto.home :

*       -fstype=nfs4,rw,nosuid,soft,rsize=8192,wsize=8192  nfs.server.com:/home/&

my exports :

/nfs4exports 192.168.0.0/255.255.0.0(rw,no_subtree_check,nohide,no_root_squash,fsid=0)
/nfs4exports/home 192.168.0.0/255.255.0.0(rw,no_subtree_check,nohide,no_root_squash)

my idmapd.conf :

[General]

Verbosity = 0
Pipefs-Directory = /var/lib/nfs/rpc_pipefs
Domain = bb.int.threatmetrix.com

[Mapping]

Nobody-User = nfsnobody
Nobody-Group = nfsnobody

[Translation]
Method = nsswitch

I'm using CentOS 5.4, nfs versions:

nfs-utils-1.0.9-42.el5 nfs-utils-lib-1.0.8-7.6.el5 nfs4-acl-tools-0.3.3-1.el5

We are using RHEL5.3 with a Clarion SAN attached by FC.

Our situation is that we have a number of LUNs presented to Hosts and we want to dynamically present the LUNs to Xen Guests. We are not sure on what the best practice approach is to set this up. The Xen guests will form a cluster together and need the LUNs only for data partitions, i.e. when they are actively running services.

So one approach would be to always present all disks to all Xen guests, and then rely up on the cluster software, and mount itself, to not mount the disk twice in two locations. This sounds kinda risky and also is not very secure (one cracked guest can see/destroy all the data).

Another approach would be to dynamically add and remove the disks from the Xen guests at the dom0 level (using xm block-attach). This could work but sounds slightly complicated, I'm wondering whether Red Hat Cluster Suite supports this in some way or whether there are scripts to do this.

Yet another approach would be to have the LUNs endpointed at the Xen guests themselves - I'm not sure whether this is technically possible since the multipathing has to be done at the Host level.

While performing an upgrade, our in-house software started to report the following strange error. It is a C++ application talking to a remote PostgreSQL database, defined through ODBC:

ODBC error state S1092, native error 0. 
[unixODBC][Driver Manager]Invalid attribute/option identifier

Both the client and the server are Centos 5.4 Xen guests with the following RPMs installed:

postgresql-libs-8.1.18-2.el5_4.1
postgresql-odbc-08.01.0200-3.1
postgresql-8.1.18-2.el5_4.1
postgresql-server-8.1.18-2.el5_4.1

Its possible the schema changed as part of the upgrade, could this explain the error message? What does this error message actually indicate, and do you know any likely causes of it?

We have DRBD volumes that are presented to guests, and we often snapshot them with the intention of reverting back to the snapshots. I understand in LVM, snapshoting then working with the snapshot is preferred when doing something experimental but often we make the snapshot simply as a backup to revert to if something goes wrong (known good state).

Mounting the disks and rsyncing is possible but sometimes the contents is complex with multiple partitions (or LV's inside the disk). Using dd to copy all the data is also possible but it is very time consuming. What I was more hoping to do is can I use lvconvert and convert a snapshot into a normal linear lv? Can I mirror a snapshot then use that as a linear lv?

Thanks

I have some sqlite version3 db3 files I copied off a live running production system (I know bad sysadmin bad sysadmin) for various reasons. Is there some sqlite command I can run that will verify that all the data can be read out of these files (I don't mind if it takes a while).

I was considering hacking up some perl which dumps out all data and then re-imports it into new files. I think sqlite will throw an exception if it encounters corrupt data. Is there a better way?

I'm CentOS 5.3 and sqlite-3.3.6-2