This repository has been archived by the owner on Jan 31, 2020. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 8
/
INSTALLATION_REQUEST_PROTOCOL
40 lines (32 loc) · 2.56 KB
/
INSTALLATION_REQUEST_PROTOCOL
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
1. Initial requests for analysis software go to the apipe-support queue.
2. If the software is an update to something already used in a pipeline, or is ready immediately to go into a pipeline:
2.1 The ticket moves into the apipe-install queue.
2.2 APIPE fire team does packaging for the cluster.
3. For requests NOT directly to be used in a pipeline:
3.1 APIPE move the ticket into hardware as before.
3.2 APIPE adds a line to the unpackged-non-pipeline file in the genome-snapshot-deps package.
3.3 Rich or others in User Support perform their traditional role installing to /gsc/pkg/bio.
3.4 After installation, the ticket goes into the apipe-install queue, with a status of "stalled".
3.5 Analysts use the software.
4. The APIPE fire team identifies apipe-install tickets from the above which are "stalled" and > 60 days old.
4.1 They check with the analysts to see whether the software is still being used.
4.2 If it is NOT in use:
* APIPE moves it from /gsc/pkg/bio into another location which is inaccessible to the blades and home directories, (/gsc/pkg/bio-archive?).
* It is removed from the list genome-snapshot-deps/unpackaged-non-pipeline list and added to genome-snapshot-deps/unpackaged-abandoned.
4.3 If it IS in use, and we know it needs to go into a pipeline:
* The ticket is moved out of the stalled state.
* Developers working on that pipeline do the packaging.
* It is removed from the genome-snapshot-deps/unpackaged-non-pipeline list, and is put into the appropriate package list.
4.4 If it IS in use, but still not in pipelines, just ad-hoc analyst tools:
* APIPE simply resolves the ticket.
* Packages lists are left as they were.
5. Later, if the tool under /gsc/pkg/bio needs to be run from a pipeline:
5.1 The developer digs out the stashed code from /gsc/pkg/bio or bio-archive.
5.2 We begin, with a head start, on packaging, using the above.
5.3 We retire the version under /gsc/pkg/bio as described under 4.2 above.
In Summary, there are three new lists in the genome-snapshot-deps package which track all new unpackaged software.
From the time a request comes in for software it is put in one of these lists, and it only leaves this set of lists when it is actually packaged.
* unpackaged-pipeline Old things the pipeline uses from /gsc/pkg/bio (never grow this)
* unpackaged-non-pipeline New things in /gsc/pkg/bio not used in pipelines
* unpackaged-abandoned When we stop using something in /gsc/pkg/bio and hide it, list it here.
For details on the lists of packaged software, see the general README in the genome-snapshot-deps package.